1 AN ANALYSIS OF TRANSPORTATION DEMAND PATTERNS IN GHANA By MARIAN ANKOMAH A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2019
2 2019 Marian Owiredua Ankomah
3 To my awesome family
4 ACKNOWLEDGEMENTS The dissertation would not be possible without the immense support from my family, acad emic advisor, church and friends despite the arduous doctoral journey. First, I would like to thank my advisor, Dr. Siva Srinivassan, for mentoring me in my qualities. Durin g the past four years of my doctoral journey, he has been a for his guidance and encouragement especially during the times when I had to change my dissertation topic due to data inaccessibility concerns. I appreciate his commitment to supporting my professional development and pushing me to participate in workshops and other programs to build me up professionally. He also cared about my personal development and could easi ly resonate with my needs as an international student since he was once like me. I acknowledge his investment in me and giving me the opportunity to grow in conducting scientific research. ; Dr Lily Elefteridou Dr. Ruth Steiner and Dr. Tia Mang for their assistance and for providing invaluable comments and feedback during my dissertation defense. I would like to thank my colleagues at the University of Florida Transportation Institute (UFTI) for their support. who provided me with contacts in Ghana for the data needed for my dissertation. Abraham, who also served as my church leader pushed me to reach my goal of completing my doctoral degrees within four years and offered unselfish advice during the final stages of my dissertation whiles job hunting.
5 I am very grateful for the wonderful community that I was surrounded with throughout my journey in Gainesville. Special thanks go to my Church of Pentecost Incorpo rated Gainesville Assembly church family especially Mr. and Mrs. Tackey Otoo, Mr. and Mrs. Yarney, Mr. and Mrs. Danso, Richard Boampong, Pamela Deane, and Joyce Dankyi for their relentless support in various forms. I wish to acknowledge my loyal roommate, Shirley Tandoh who has been a for me during the times I fell sick and when I underwent eye surgery in 2015. To the lovely kids (Eyram, Sarah, Ryan and Mikayla) who I was privileged to babysit someti mes, I appreciate the balance you brought into my life in helping me to prioritize and appreciate the value of my time. I want to express my gratitude to Joyce Assan for the many times she surprised me with free groceries an d her constant care for me. Additionally, I would like acknowledge these wonderful friends who have been there for me in several ways by ensuring that I was never sad. I would like to thank Naa Nortey and her family who adopted me as their family member a nd supported me always. To Rebecca Mensah, Yvonne Afriyie, Catherine Ikponwomba, and Maame Adwoa Sakyi Appiah, I appreciate your unwavering support, prayers, and advice during this journey. I am grateful for all the youths and young adults of Jesus People North America (JPNA) who were a great source of encouragement and motivation to me, especially recognizing that many youths in the church looked up to me. Most importantly, I would like to thank my family for their endless support, encouragements and praye sacrifices especially in meeting my financial needs and just making sure that I was
6 comfortable. I appreciate their advice and encouragement in pushing me to pursue my education to the highest level. I would like to thank my brothers and their families for always being there for me. I appreciate the many times they went the extra mile to make youngest brother, Fre derick Ankomah who was my source of strength and encouragement.
7 TABLE OF CONTENTS page ACKNOWLEDGEMENTS ................................ ................................ ............................... 4 LIST OF TABLES ................................ ................................ ................................ ............ 9 LIST OF FIGURES ................................ ................................ ................................ ........ 10 ABSTRACT ................................ ................................ ................................ ................... 11 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 13 Motivation ................................ ................................ ................................ ............... 13 Traffic Congestion ................................ ................................ ................................ ... 14 Safety ................................ ................................ ................................ ...................... 16 Summary ................................ ................................ ................................ ................ 17 Research Organiz ation ................................ ................................ ........................... 18 2 TRIP GENERATION OF WORKING AND NON WORKING ADULTS IN GHANA ................................ ................................ ................................ ................... 21 Literature Review ................................ ................................ ................................ .... 21 Data ................................ ................................ ................................ ........................ 24 Data Structure ................................ ................................ ................................ .. 24 Data Descriptives ................................ ................................ ............................. 27 Modeling Methodology ................................ ................................ ............................ 30 Results ................................ ................................ ................................ .................... 32 Working Adults ................................ ................................ ................................ 32 Non working Adul ts ................................ ................................ .......................... 35 Summary and Conclusion ................................ ................................ ....................... 36 3 VEHICLE OWNERSHIP MODEL ................................ ................................ ............ 59 Literatur e Review ................................ ................................ ................................ .... 59 Data ................................ ................................ ................................ ........................ 62 Data Structure ................................ ................................ ................................ .. 62 Data Descriptives ................................ ................................ ............................. 63 Modeling Methodology ................................ ................................ ............................ 65 Results ................................ ................................ ................................ .................... 66 Summary and Conclusion ................................ ................................ ....................... 69 4 ................................ ..................... 76 Literature Review ................................ ................................ ................................ .... 76
8 Data ................................ ................................ ................................ ........................ 79 Data Structure ................................ ................................ ................................ .. 79 Data Descriptives ................................ ................................ ............................. 80 Modeling Methodology ................................ ................................ ............................ 81 Results ................................ ................................ ................................ .................... 83 5 SUMMARY AND CONCLUSION ................................ ................................ ............ 95 Summary ................................ ................................ ................................ ................ 95 Research Implications ................................ ................................ ............................. 97 LIST OF REFERENCES ................................ ................................ ............................. 100 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 105
9 LIST OF TABL ES Table page 2 1 Characteristics of Households ................................ ................................ ............ 41 2 2 Male and Female Proportion of Trips in the Last Seven Days per Mode ............ 43 2 3 Characteristic s of Working Adults in Ghana ................................ ...................... 46 2 4 Characteristics of Non Working Adults i n Ghana ................................ ............... 50 2 5 Trip Generation Model for Working Adults in Ghana (Vehicles Owned) ............. 54 2 6 Trip Generation Model for Working Adults in Ghana (Foot, Taxi, Bus) ............... 55 2 7 Trip Generation Model for Non Working Adults in Ghana (Vehicles Owned) ..... 57 2 8 Trip Generation Model for Non Working Adults in Ghana (Foot, T axi, Bus) ....... 58 3 1 Characteristics of Households ................................ ................................ ............ 72 3 2 MNL Model on Vehicle Ownership ................................ ................................ ..... 74 4 1 Bus as a Primary Mode by Activity ................................ ................................ ..... 86 4 2 Proportion of the Perception of Adults on Public Transportation (Bus) Based on their Activity Purpose ................................ ................................ ..................... 86 4 3 Descriptive Statistics of All Adults who have Ever Used a Bus .......................... 87 4 4 Descriptive Statistics of Adults who have Ever Used a Bus based on their Overall Satisfacti on ................................ ................................ ............................. 89 4 5 Binary Logit Models on Bus Perception ................................ .............................. 91 4 6 Binary Logit Models on Overall Satisfaction with using Bus ............................... 93
10 LIST OF FIGURES Figure page 1 1 Maps of Ghana ................................ ................................ ................................ ... 19 1 2 Motor Vehicle in use from 2005 to 2015 for Ghana ................................ ............ 20 2 1 Reclassified Education System ................................ ................................ .......... 39 2 2 Data Assembly Structure for Chapters 2 and 3 ................................ .................. 40 2 3 Frequency Distribution of the Total Number of Trips in the Last Seven Days per Mode ................................ ................................ ................................ ............ 43 2 4 Frequency Distribution of the Total Number of Foot Trips on a Normal Day ...... 44 2 5 Male and Female Proportion of Foot Trips on a Normal Day ............................. 44 2 6 .................... 45 2 7 Proportion of Non ............ 45 3 1 Proportion of Vehicle Fleet Ownership by Regional Location ............................. 71 4 1 Perception ranking scale ................................ ................................ .................... 86
11 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of t he Requirements for the Degree of Doctor of Philosophy AN ANALYSIS OF TRANS PORTATION DEMAND PAT TERNS IN GHANA By Marian Ankomah August 2019 Chair: Siva Srinivasan Major: Civil Engineering Understanding the travel behavior of road users is important in the development of transportation plans. In Ghana, just like most developing countries, there appear to be few studies on trip generation or vehicle ownership as most studies focus on mode choice modeling. This research focuses on understanding the travel behavior of road users by examining the vehicle fleet composition of a household and the trip generation of both working and non working adults in Ghana. The study also performs a multivariate analysis on the perception of adults on the use of public trans portation. This study utilizes data from the 2012 Ghana Transport Indicator Database Survey which was conducted by the Ghana Statistical Service. The trip generation by mode of travel (walk, bicycle, motorcycle, car, taxi, and bus) for working and non wor king adults are investigated through developing negative binomial regression models. Findings from the study show that gender, age, education level, household income, and residential location all show a significant impact on the number of trips generate d f or the different trip modes. From the results, the marital status of an adult significantly affects the car, bicycle, motorcycle and taxi trips a worker make ; and the car, motorcycle and bus trips of a non worker. Both workers and non
12 workers with at least households that reside in the city center make fewer trips by car than similar adults that live outside the city center. Results from the vehicle ownership multinomial logit model show t hat high income households are more likely to own at least one type of vehicle. However the effect of the household residential location significantly impacts t he kind of vehicle a household is likely to own with the exception of households that own at least one car. The public transport user s perceptions are examined by developing binary logit models for different satisfaction levels. The common finding s from the models indicate that, the more adults use the bus for different activities, the more the y adjust to the bus conditions, availability, schedule and route and as such become more comfortable and satisfied with them.
13 CHAPTER 1 INTRODUCTION Motivation Travel demand forecasting plays a significant role in the development of transportation plans a nd evaluation of transportation infrastructure. Travel needs/demands keep on changing as such it is imperative to make plans that account for the possible future changes. In developing countries, especially Ghana (Figure 1 1) there is a growing need for a transportation system that captures the growing needs of its population. In Ghana, land use planning also conflicts with transportation planning. For example, residential areas being used for commercial purposes. Also, the vehicle population grows higher than the transportation network as such increasing pressure on the transportation infrastructure. The vehicle population growth is largely due to high population growth, urbanization (Figure 1 1 ) and a sense of vehicle ownership. There is a sense of pride and achievement in owning a vehicle and the notation that individuals who usually patronize public transport are from poor or low income households. Therefore, households who own vehicles are rarely seen using public transport as that diminishes their soc ial status. As of 2015, there were 1.28 billion vehicles in the world, with a projected vehicle growth of 2 billion by 2040 (Smith 2016 ). According to UNECE (2017), the majority of this growth is reflected in developing countries with a large proportion being used cars that will be imported from developed countries. Ghana charges additional duty on vehicles older than 5 years old but that does not deter people from importing older than 5 years' vehicles into the country due to the growing need for owning a vehicle. People even import accident vehicles with vehicle parts into the country which are rebuilt
14 cheaply and sold cheaply in the market. Ghana remains in the top 5 of importers of used vehicles with the USA being the main exporting country. Over 85% of used cars from the USA are exported to West Africa with Ghana and Nigeria accepting a larger proportion of these vehicles (Baskin 2018 ). Vehicle growth is not only increasing on our limited road infrastructure but the larger proportion of these vehicle s being second hand is important in understanding how the increasing vehicle growth impacts travel demand considering safety, and air quality. In fact, of the new car sales in sub Saharan Africa, 80% of them are from South Africa with the remaining 20% spr ead in the rest of the African countries. In Ghana, the total number of registered vehicles by 2015 was 890,000 vehicles that is 3 out of every 100 persons in Ghana owing at least one vehicle. Figure 1 2 below shows the vehicle in use from 2005 to 2015. A utomobile ownership in Ghana increases by averagely 40,000 vehicles per year (OICA). A more significant proportion of the vehicles (60%) owned in the country are in Accra, the capital city and business hub of Ghana whose vehicle ownership is expected to ri se to over 1 million by 2023 ( Fiagborlo 2017 ). About 15% of personal cars in use in Accra use 85% of the road space as such the need to develop strategies to help shift the dependency of the car owners using their cars to some form of public transport. Th is lead to the introduction of Bus Rapid Transport on the roads of Accra in 2016 to provide this modal shift (Adogla Bessa 2016 ). Traffic Congestion Vehicle ownership is a crucial contributor to traffic congestion. The issue of traffic congestion remains one of the significant problems of transportation in Ghana, especially in urban areas. The Increase in vehicle ownership has aggravated the issue of traffic congestion in such a way that increases in the capacity of road infrastructure
15 doesn't even help to curb this problem (Hart, 2016). Meanwhile, In Hart (2016)'s book on "Ghana on the Go," they identified that expansion of the vehicle ownership market in the 1990s in Ghana brought about an increase in driving schools creating more jobs and improving the e conomy. The Increase in vehicle ownership also posed a potentially significant increase in road traffic crashes. Also, despite the significant increase in vehicle ownership especially private cars, a large proportion of residents in the urban areas rely he avily on commercial transport. Baffour (2010) establish ed the prevalence of traffic congestion to be during the AM and Peak hours in the country and g ave the road facilit ies in Accra a level of service F during such hours. The author expressed his uncertai nty of Accra having a Traffic Demand Forecasting Model. Further, the article identifies poor coordination and signal timing, inefficiency and frequent breakdown of traffic control devices in addition to poor road designs as other cause s of traffic congesti on in the capital city ( Baffour 2010). Traffic control devices not appropriately stationed at the right intersections also contribute to traffic congestion. Usually, the police assume a position in areas of poor traffic coordination or malfunctioning to h elp curb the congestion temporarily. In an article by Mordy (2017), traffic congestion decreases productivity and causes loss of billions of cedis annually of productive hours to the country. Low productivity at work usually stems from stress from enduring long hours of delay on the road. Traffic congestion doesn't only arise because of an increase in vehicle ownership, but over 50% of the roads in Ghana are bad and untarred. This is primarily due to the government to fix the roads in the countr y and complete existing road projects due to insufficient funds. Furthermore, the rough roads also pose safety implications in
16 addition to traffic congestion. Hmm (2011) Identifies the leading cause of traffic congestion as the road infrastructure's capaci ty inability to meet the daily traffic needs, road construction, and traffic accidents. Traffic congestion indirectly affects the health of the public because of environmental and health pollution. The article also states that building more roads to meet t he increasing vehicle need of the country is not the surest way of solving the issue of traffic congestion and other transport problems thus admonishes engineers and planners to implement other plans and strategies of effectively managing traffic congestio n. Knowing the vehicle ownership of the households and restricting the number of vehicles a household uses was one of the traffic management measures that the author identified could be implemented. Safety Road crashes still pose as one of the leading ca uses of death in Africa and the world at large. It is seen to be the leading cause of death among young people between the ages of 15 and 29. Also, middle and low income countries constitute 90 % of the global road crashes showing double fatality rates as high income countries. (WHO, 2015) All African countries fall under the umbrella of low to middle income countries with Ghana classified as a lower income country according to the World Bank. Therefore, in as much as road crashes is a global concern, it is of most significant concern to Ghana. Generally, road crashes have been increasing over the years with an increase in the number of road traffic fatalities despite the reduction in the fatality rate in Ghana. (Auditor General, 2010). Also, the report by C oleman (2014) and Nyamuame, Godwin Y., et al. (2015) supports the claim that the total number of road traffic fatalities has been increasing progressively since 2000. The National Road Safety Strategy (NRSS) set up an objective to reduce road traffic fatal ities to less than 1000 by 2015.
17 However, the 2016 report from the National Road Safety Commission (NRSC) showed that the total number of road crashes has been decreasing although the number of fatalities increased in 2016. The country is still battling th is safety issue with the increase in vehicle ownership despite the many efforts that the NRSC has put in place in ensuring road safety. Interestingly, vehicle types are gradually being recognized as a salient factor in causing road accidents as seen in the study by Mohammed et al. (2015). The studies show that vehicle ownership and the travel behavior of road users significantly impact the safety of road users, therefore, the utmost need in understanding how trip generation makes us understand the travel pa ttern of road users which consecutively influence where road traffic crashes will often occur. Summary There is a critical need to develop improved travel demand forecasting models for Ghana as the current forecasting model used in practice is an aggregate model with few explanatory variables. Demand forecasting models typically comprise of two main components: long term choices such as residential location and vehicle ownership and short term choices such as trip frequency, mode and time of day. This study contributes first to understanding the factors influencing vehicle fleet ownership and developing a trip generation model for Ghana which relates individual and household characteristics including the household vehicle fleet ownership levels, residential location, and other perception explanatory variables to the number of trips by mode. Additionally, this study also seeks to understand the public transportation perceptions of road users in Ghana.
18 Research Organization Chapter 2 presents a liter ature revie w, data description, methodology and model results on trip generation in Ghana Chapter 3 presents a literature review data description, methodology and model results on household vehicle fleet composition in Ghana. Chapter 4 presents a literature review, data description, methodology and models result i n understanding public transport perceptions. Chapter 5 provides a summary of the studies and recommendations for future research.
19 (A) (B) Figure 1 1 Maps of Ghana A) Regional Boundaries. B) Proportion of Urban Population 1 1 Songsore, Jacob. "The urban transition in Ghana: Ur banization, national development and poverty reduction." University of Ghana, Legon Accra (2009).
20 Figure 1 2 Motor Vehicle in use from 2005 to 2015 for Ghana 0 100 200 300 400 500 600 700 800 900 1,000 2004 2006 2008 2010 2012 2014 2016 Motor Vehicles Use (thousands) Year
21 CHAPTER 2 TRIP GENERATION OF WO RKING AND NON WORKING ADULTS IN GHANA Literature Review This section discusses the literature synthesis on trip generation in developing countries. The trends of trip generation are conventional and well developed. In the United States, it is done by purp ose as trip generation captures the reason behind a trip is being made. An example of how trip generation by purpose is done in the US is reflected in the studies by Lim and Srinivasan (2011). Their study focused on comparing four different econometric str uctures (linear regression, log linear model, negative binominal model, and ordered probit model) for trip generation modeling for three trip purposes home based work (HBW), home based other (HBO) and non home based (NHB). The data source for their study was from the 2001 and 2009 U.S. National Household Travel Surveys. The 2001 data was used to estimate the trip generation models, and the 2009 data was used to validate the estimated model. Their findings recommend the use of ordered probit models against the traditional regression models. In Ghana and most developing countries, there appear to be few studies on trip generation, however, there is a bunch of studies that have looked at mode choice. For example, Abane (2011) study the attitudes and travel be havior of Ghanaians from four metropolitan areas Accra, Tamale, Kumasi, Sekondi Takoradi. Their study shows that mode choice is significantly impacted by the affordability and availability of those modes. This study is supported in the paper by Agyemang (2017) who identifies perceptions of convenience (that can be related to affordability and availability) as a and persons with higher levels of education were less likely to travel by trotro than a car,
22 and this result is consistent with the finding by Ye and Titheridge (2017), and Ashalatha et al. (2012). In addition to younger persons being more likely to prefer bus than cars, Ashalatha et al. (2012) findings from studying the mode choice behavior of commuters in Thiruvananthapuram, India show that, as age increases, the preference of cars increases with a decreasing preference for two wheelers compared to bus es Also, gender significantly influences the choice of a particular mode as male persons are more likely to choose a car or two wheeler over a bus. A well established finding which was also identified by them was that persons with higher monthly incomes are likely to prefer cars. The socio economic characterist ic of a household does truly impact their mode choice. Meanwhile, in understanding the trip characteristics that influence households with zero vehicle ownership, Arasan et al. (1996) identified the trip distance as the key factor impacting mode choice. In comparing the mode choice of employees in the formal sector in the two largest cities in Ghana, that is Accra and Kumasi, Abane, Albert (2011) and Amoh Gyimah and Nimako Aidoo (2013) found that high income workers are less likely to use public transport. Furthermore, findings from Amoh Gyimah and Nimako Aidoo (2013) show that workers with households of family size greater than two are less likely to choose non motorized transport. From their study, the distance from home significantly affects mode choice a s workers are less likely to travel to work by public transport for distances greater than 5km. Etminani Ghasrodashti and Mahyar (2015) presented an interesting finding on how different lifestyles of a person influences their mode choice for non working tr ips. From their study, they identified that people with Modern Oriented Lifestyle (MOL) are more likely to make more non working trips by car than persons with
23 Consumer Oriented Lifestyle (COL) who are more likely to make non working trips by transit. Also persons with more traditional and educational oriented lifestyles are more likely to make non working trips by walking or cycling Soltani et al. (2011) in a bid to examine the influence of urban physical characteristics, socio economic characteristics an d public transport characteristics on trip generation in Shiraz, Iran, identified the socio economic status to be the most significant factor impacting trip generation. From their study, the trip makers' socio economic characters (age, employment status, e ducation, migration, property value) which were aggregated as one variable showed that an increase in the level of these socio economic statuses would significantly increase trip generation. Furthermore, in terms of trip generation by purpose, Takyi (1990) employed the cross classification method to determine household characteristics and its influence in the trip making. Results from the study significantly show that an increase in the household size will increase the trotro and walking trip rate for work, shopping, and school trips. Also, Afolabi et al. (2017) show a relationship between commuter's frequencies of travel by their level of income. In Durban, South Africa, Venter et al. (2007) examine how gender differences together with the residential loca tion of low income urban househol ds influence travel behavior. Households in low income areas with limited schools, jobs, and social amenities generally made more trips by car and public transport than walking. Also, their study identified travel by taxi a s a popular mode of travel. Per their study, three fourths of low income households use taxis as their travel means to/from work with the use of taxi occupying 50% of the mode share for other trip purposes. Their findings on
24 households in the periphery of the urban areas supported previous results on women making fewer trips of longer distances than men usually by walking or using public transit. However, both women and women in urban areas show similar mobility patterns. Further studies by Salon and Gulyan i (2010) elucidate how poverty affects the transport choices of slum households in Nairobi, Kenya. First, they acknowledged that poverty stricken slum households do not have more than one mode option as most of these residents almost always walk. They deve loped a logit model identifying the factors that influence slum children's choice of walk ing to school away from their home and an MNL model comparing the work travel choices of the men and women in the slum considering two choice variables of walk or mata tu (bus). Their findings showed poverty to be a significant factor in the choice to ride a matutu by all workers. Men from slum households with children were more likely to commute to work by walking or riding a matutu than working in their home settlement Affordability of the matutu also largely influence s the slum's residence choice of patronizing that mode. Additionally, reduced fares of the matutus or improved economic status of those slum residents will significantly increase the use of matutu. Overal l, there has been an appreciable study on mode choice in developing countries but very few on trip generation. Therefore, this study aims to model the trip generation by mode of working and non working adults in Ghana and the vehicle fleet ownership of all households. Data Data Structure The data described in this section is used for the analysis performed in Chapter s 3 and 4. This study was performed using data from the 2012 Ghana Transport Indicator
25 Database Survey which was conducted by the Ghana Stat istical Service. This data which represents the second phase of the first ever nationwide household based transport survey was conducted over a 3 month period between September 2012 and December 2012. The survey was administered to 6000 households which re present all the households in Ghana from all the ten regions. The vehicle ownership study considers all the households but the analysis on trip generation focus es only on all working and non working adults in Ghana. The data includes some basic socio econo mic characteristics of the individuals of a household (such as age, gender, ethnic educational background (such as highest educational level, current grade, and mean s of transport to and from school). Also, the data provides some socio economic characteristics of the household characteristics such as size and income, and some location variables such as region and district where the household lives. The survey also col lected data on the various transport activities of individuals in the household concerning health, economic activities, market activities, and other household activities. The detail characteristics of the individual and households' characteristics make the data viable in understanding the travel behaviors of households with respect to mode choice and help generate a travel behavior model for the nation. The data contains individual and household variables for 23238 individuals from 6000 households' selectio n from a total of 400 Enumeration Areas (EA). However, the data doesn't explicitly provide any variable that identifies each household member belonging to the same household. Therefore, with supplemental information on the
26 enumeration areas listing that wa s used for the survey which was provided by the Ghana Statistical services, a unique Household ID code was generated to match the household members with their respective households. The data provides a variable as household number (HHID); however, a freque ncy distribution of this variable showed that there were 20 household number s with 15 of them each having more than 1450 households. From a further probe into this variable, we were able to identify that the household numbers referred to the number (starti ng from 1) assigned to each household that was randomly selected in each enumeration area for the interview. Therefore, a unique code combining each HHID in their respective EA was generated. For example, a code of 2312 refers to an EA of 231 and an HHID o f 2). With this code generated, household members were linked to their respective households and we were able to generate some household characteristics such as household size, and the total number of adults, children, workers and students in a household. Out of the national sample of 23,238, 58.4% of them representing 12734 were adults (that is 18 years or older). Further basic statistics of this study are based on the sample size of all adult household members. In preparing the data for analysis, the data was subject to a thorough cleaning and some data variables were reclassified and adjusted for clarity. There were some inconsistencies in the data sample, therefore some cases were filtered out to help clean the data. Some data cases showed a value of zer o for the average travel time to/from school, work, health facility, and market; therefore careful check was done before filtering out the zero values in this variable. For the school related purpose, the average travel time of zero was filtered off for ev ery transport means except boarding school. It
27 is assumed that there is no major travel activity from a residence or any origin to a school if it is a boarding school. For health and market, the average travel time of zero for every transport mode was filt ered off. In terms of work, 1863 of the cases had an average travel time of zero. However, a crosstab of average travel time to/from work with the variable: "does work require travel from residence," revealed that the time of zero was stemming from househo lds that required travel from residence to work. A cross tabulation of the average travel time with the variable: "means of travel to work" shows that the average travel time of zero is predominantly by traveling by foot thus making it difficult to explain Therefore, the 1863 cases were kept for this analysis purpose. The different modes of transport for each purpose were also reclassified. All trips done by a shared public taxi or individual taxi was reclassified as a taxi based on purpose. All means of t a boat/canoe/ferry and train. Therefore, the primary means of transport identified in this data are taxi s bu s es private cars, motorcycles, bicycles, foot, and others. Due to the constant evolution of the education system in Ghana, the highest grade variable of the members which represents the highest educational level for non students in the household were rec lassified into a standard educational sy stem for easy comprehension ( Figure 2 1). The overall data assembly structure for Chapters 2 and 3 is presented schematically in Figure 2 2. Data Descriptives Thorough cleaning and consistency checks generated a cle an sample of 22784. The full sample shows a gender distribution of 51.6% female and 48.4% male with an
28 average age of 24.56. On average, there are about 3.99 persons per household with 2.17 adults per household and 2.05 workers per household. The workers c omprise of employees, employers, self employed workers, unpaid family worker s and apprentices. About 68.2% of households have children with an average of 2.67 children per household. With respect to gender, 3 5.1 % of the households' adults are predominantl y women, 26.8% are predominately men with the remaining representing households with an equal proportion of male and female adults. In terms of location, 42. 7 % of the households live in urban areas whereas 57. 3 % live in rural areas. The topmost 5 regions c apturing the households' locations are Ashanti Region (18. 8 %), Greater Accra Region (14 .3 %). Northern Region (1 0 .4%), Eastern Region (11. 4 %) and Brong Ahafo Region (9. 8 %). In terms of culture, 46.4% of the households are from the Akan ethnic group, 19.5% are Mole Dagbanis, 13.3% are Ewes, and 6.4% are Ga/Dangmes with the remaining 13.6% belonging to the other ethnic groups. The highest education level of all non schooling persons in the household was also considered with only 3.7 % of them having a bachelo r's degree. The descriptive statistics also show the monthly income of each household, the residential location of the household, marital status of the household person, religion and some perception variables. Further, in terms of the households vehicle ow nership, 23 .7 .1 % own motorcycles, 4 .2 % own cars, 0.3% own busses, and 0.4% own trucks. The detailed descriptives of the household characteristics are presented in Table 2 1 This study focuses on both working and non wo rking adults only therefore, F igure 2 3 shows the frequency distribution of the total number of trips by car, bus, taxi,
29 motorcycle and bicycle made by both working and non working adults in the last seven days from the day the respondent completed the su rvey. The distribution shows a significant number of zero trips for all the different transport modes. Adults make more taxi and bus trips and there is no significant difference in the number of trips by mode made both males and females. Meanwhile, males a re seen to generate more trips by motorcycle and bicycle than females as shown in Table 2 2 For trips on foot, the number of trips the adult makes on a typical day was recorded as such Figure 2 4 shows the frequency distribution for that. In both figures, the number of trips after the 40th trip (which represents 0.1% of the sample) was aggregated with the 40th trip. Both males and females show similar foot trip pattern ( Figure 2 5 ) Figures 2 6 and 2 7 also provides a map view of the proportion of working and non working adults respectively by trip mode based on their metropolitan location. Table 2 3 show s a brief description of the key variables considered in this analysis describing the travel behavior of the various household working adults. These tables present personal and household descriptive statistics of the variables describing the trip generation for working adults. From the data, 48.4% of the household members were actively involved in a paid job/work with almost all the worke rs (99.8%) working in the private sector. However, 47.8% of the households worked in the last 12 months. About 41.6% of the workers in Ghana require some form of travel to work from their residence. The data descriptive tables are organized as follows: the descriptive statistics for worki ng adults in Ghana first (Table 2 3 ), followed by the descriptive statistics for non working adults in Ghana (Table 2 4 )
30 Modeling Methodology This section of the study discusses the modeling methodology used for analysis. In understanding how household characteristics, person characteristics, residential location, and vehicle ownership impact the frequency of trips made by a working adult, the Negative Binomial Regression (NBR) was e mployed. Although the Linear for over dispersed data and underestimates the true variance. Trip frequencies are skewed and differ from locations. Some locations based on certain characteristics may generate more trips by a certain mode than others, whereas some locations may generate no trips by that same mode. Trip frequencies are over dispersed. The data (F igure 2 3 ) shows a large proportion of the household members making zero trips, therefore, in a bid of capturing this over dispersion and excess zeros in the number of trips variable, other methodologies for modeling this count data was considered: the Zero Inflated Poisson (ZIP) Regression and the Negative Binomial Regression. Although the ZIP model attempts to account for the excess zeros in the data, it was important to understand the reason for generating zero trip counts. Even though the question about the number of car, moto, bike, bus, taxi and foot trips made was asked to everyone, it does not suggest that everyone made those trips. Therefore some of the possible reasons for a household member to report a zero count (trip) for a particular mode was looked at. Was it because this member didn't own the vehicle f or that trip mode or didn't have to make any trip at all (stay home member?) or only some modes were available? Since it was quite difficult to decipher the processes that generated zero trips, we used the Negative Binomial Regression (NBR) for this analys is. With an over dispersed data with a large variance than the mean the NBR
31 was ideal for this analysis The NBR model developed examines a relationship between the number of trips a working/non working adult makes for all trip purposes based on a set of explanatory variables. The NBR is a type of the generalized linear models for modeling count ( dependent variable trip frequency) with its variance larger than the mean. Therefore, the variance of a negative binomial general form is a function of the mean as shown below: Where The NBR models were developed using the STATA statistical software and the models were subject to seve ral iterations to obtain our best models. A model was developed first with all the explanatory variables and the highly insignificant ones were filtered off. Initially, an 85% confidence was considered to garner the strongly correlated variables. Then the explanatory variables with a 90% confidence were kept to be tested for during the intermediate model specification. Finally, after several iterations, the best model was built with explanatory variables showing at least a 95% confidence. The best model cap tures only the significant variables that show a strong correlation between the explanatory variables and their impact on trip generation. Six NBR models were developed for all the trip purposes considered (car, bicycle, motorcycle, foot, taxi, bus) for bo th working and non working adults in Ghana. Therefore, a total of 12 NBR models were built and analyzed.
32 Results In this section, we present the results of the negative binomial regression model with the coefficients of the estimate and their t statistics All the explanatory variables explain their impact on the number of trips made. Tables 2 5 through 2 6 presents the trip frequency models for working adults in Ghana, and Tables 2 7 through 2 8 shows the non Working A dults The neg ative coefficients of the household size on trips made by car, foot, and taxi show that the number of trips decreases with increasing household size for car, foot, and taxi trips. From the results, workers in households with more adults make fewer trips by bicycle and motorcycle. Also, more workers in a household are less likely to make bus trips. The gender composition of a household also plays a significant role in the trip generation by car, bicycle, motorcycle, and taxi. Female adult workers make fewer trips by these modes than male adult workers. The significant negative coefficient with high t stats values on the gender variable for bicycle and motorcycle show its strong determinant in the trip generation by bicycle or motorcycle. The age of an adult w orker has a significant impact on trip generation by car, bicycle, motorcycle, and foot older workers make more trips by car and fewer trips by foot, bicycle, and motorcycle. This is consistent with the findings from Agyemang, Ernest (2017) on an older p erson traveling more by car. Education level plays a significant role in trip generation and mode choice in general. Workers with a bachelor's degree s make most trips by car, taxi, and bus than workers with lower educational levels. Findings by Amoh Gyimah and Nimako Aidoo (2013) also identified that households with a diploma degree or higher make more car
33 trips compared to households of lower educational status. Workers who are currently in school make more trips by bicycle and foot and fewer trips by taxi Furthermore, the results suggest that all workers with a middle school certificate or higher who do not own any vehicle (car, bicycle, motorcycle) are more likely to make more trips by taxi and bus although working adults with bachelor's degree would mak e more taxi trips. The educational level of the household impacts the number of trips its workers make by taxi. According to the data, about 85% of the taxi trips are shared, and the remaining 15% are individual trips. This could be a possible explanation of why trip generation increases irrespective of worker's educational level or household income. Additionally, workers with a household income less than 1200ghc ($624 per 2012 cedi to dollar conversion rate) make fewer car trips than high income workers. W orkers with a household income between 600.01 ghc and 1200ghc make fewer foot trips than workers with a household income less than 200ghc. Household income, which also shows a positive association with the level of education significantly impacts trips gen eration. Across all trip modes except foot and bus trips, the marital status of a worker shows significant impact on trip generation. It was interesting to find out that married working adults generate more trips by car, bicycle and motorcycle than unmarri ed workers. Also, single working adults generate fewer taxi trips and more bicycle trips. Divorced or separated working adults are more likely to make more car trips than self employed workers are more likely to make bicycle and foot trip trips whereas app rentices were more likely to make bicycle trips.
34 The vehicle ownership of a household also significantly affects the bus trips a worker make s As expected, households who own a car are less likely to make bicycle and motorcycle trips. Households who own bi cycles and motorcycles are less likely to make bus and taxi trips but make more foot trips. Also, households owning at least one truck make fewer bus trips. The higher negative coefficient on bus trips for households who own at least one truck show that wo rkers in such households make fewer bus trips than workers in households that own a bicycle or motorcycle. A plausible explanation for the significance in generating fewer trips of this could be that house holds which own trucks are more financially sound a nd would most likely resort to other modes of transport like taxi or car if they own one. In terms of residential location, adult workers that reside in the city center make fewer trips by car than similar adults that live outside the city center. Similar ly, such adults generate fewer trips by taxi except for adults that live on the farm. In contrast, such adults living in the city center generate more foot trips although the number of foot trips shows a positive correlation across all the residential loca tion variables except households that live on a farm. A conceivable reason for this is that the travel activity of such workers can be located within the city center. This result can also be linked with the finding by Haybatollahi, Mohammad et al., (2015) that people in highly dense areas of the city center often walk or bike. Working adults in the Accra Metropolitan Area (AMA) create more car trips than workers in all the other major metropolitan areas in the country. Also, the number of motorcycle and tax i trips increases whereas the number of bus trips decreases for working adults in the Tamale Metro area. This result is also
35 Tamale in his study. This is because t he main mode for travel in the northern part of the country which is captured by the Tamale Metro area is the motorcycle. Non working A d ults The structure of the household significantly influences the trip generation as non working adults in households th at own no vehicle are less likely to make foot, taxi and bus trips irrespective of the household size. Such adults make fewer foot, taxi and bus trips as the household size increases. Also, households with more adults make fewer bicycle trips, and this fin ding is expected as bicycles usually have a seating capacity of one unless a rear rack is attached to the bicycle. In terms of gender, females generate fewer bicycle, motorcycle and bus trips than males, although the high negative coefficient on motorcycle and bicycle compared with bus show that females make fewer two wheeler trips than bus trips. Age is also a significant determinant in trip generation as older adults make fewer car, bicycle and foot trips. The effect of the adult's highest educational lev el and the household's income on the trip frequencie s are as anticipated. Household members currently attending school generate more trips by all modes except car and bus. The marital status of an adult and the household residential location are rather str ong predictors of the trip generation by car. Single adults make fewer car trips and more motorcycle trips depending on their location. Adults living in the towns and suburbs generate the most car trips although adults living in the suburbs also make more motorcycle, foot, taxi and bus trips. It was interesting to find out that non working adults who live on the farms make fewer foot trips. A possible explanation to this could be that, adults who live on farms may make a trip out of their farm to transport their farm produce so may be more likely to make fewer foot trips transporting such commercial produce.
36 Across all trip modes except bicycle trips, the households that own a motorcycle, bicycle or bus show a significant impact on trip generation. Interesti ngly, non working adults that own motorcycles make more foot trips. Out of the 21% of non working adults that are seniors, (adults 65 years and above) majority of them make mostly foot trips. This could be a possible explanation of why non working adults g enerate more foot trips. This result is also consistent with the findings from Hu, Xiaowei et al., (2013) whose study on understanding the travel behavior of older adults in a developing country, show that almost half of the older adults in China choose t o walk. Non working adult households that own at least one bicycle make more motorcycle trips and fewer taxi trips. Such adults with this trip pattern most likely own motorcycles as well. Adults from households that own a bus make fewer bus and taxi trips but more car trips. A plausible explanation to this travel trend could be that households which own busses are more likely able to afford to own cars and may use their buses for commercial purposes whiles they use their personal cars for most of their trip s. In terms of the metropolitan location of the household, the only significant variable for car trips is the adults that live in AMA. Such adults generate more car trips in addition to motorcycle, foot, taxi, and bus. However, adults in the Kumasi Metrop olitan Area (KMA), a comparable large metropolitan area to AMA generate fewer motorcycle trips but more taxi and bus trips. Summary and Conclusion Although understanding the travel behavior of road users is key in travel demand forecasting, there has bee n little contribution to the modeling of trip generation in developing countries This study, however, presents a trip generation model by mode for working and non working adults using the 2012 Ghana Transport Indicator Database
37 Survey. Apart from the Ghan a Statistical Services which developed a report based on this survey, the disaggregate data has not been used for any research relating to transportation or trip generation as such. Therefore, results from this data serve as a useful contribution to the Gh ana Statistical Services, Ministry of Transport and other governmental transportation agencies in developing national policies. Of the national sample of 23,238 individuals 58.4% of them represent ing 12734 were adults (that is 18 years or older) : 9648 wo rkers and 2849 non workers. A Negative Binomial Regression (NBR) model was used to model the trip generation by car, bicycle, motorcycle, foot, taxi and bus for these working and non working adults. The results from the study show that gender, age, educa tion level and residential location all have a significant impact on the number of trips generated for the different trip modes. From the working model, the marital status of an adult also significantly affects the car, bicycle, motorcycle and taxi trips a worker make s Fo r non workers, marital status has a significant impact on car, motorcycle and bicycle trips only. Both workers and non trips and taxi trips. Also, household income which show s a positive association with the level of education significantly impacts trips generation. Lower i ncome workers make fewer car trips than high er income workers the lower the income, the less likely the worker would make a car trip. Furthermore, adult h center make fewer trips by car than similar adults that live outside the city center In accounting for the effect of metropolitan location on trip generation both working and non working adults in the Accra Metropolitan Area (AMA) create more car trips than adults in all the other major metropolitan areas in the country. In addition, non working
38 adults that live in the AMA generate more motorcycle, foot, taxi, and bus trips. However, non working adults in the Kumasi Metr opolitan Area (KMA), a comparable large metropolitan area to AMA generate fewer motorcycle trips but more taxi and bus trips.
39 Figure 2 1 Reclassified Education System Highest Grade Less than middle school graduate Preschool, P1 P6, JSS1 JSS2, M1 M3 Middle school graduate/BECE JSS3/M4 Some High School SS1 SS2, S1 S4 Vocational training voca/tech/computer/comm/agric SSCE certificate/ A level certificate SS3,SS4, S5,A level Training College/Polytechnic Teacher training/nursing/polytechnic Bachelors university Other Other tertiary/other
40 Figure 2 2 Data Assembly Structure for Chapters 2 and 3
41 Table 2 1 Characteristics of Households Socio Economic Variables Percent Frequency Household size structure Single Person HH 20.3% Two Person HH 13.3% Three Person HH 15.1% Four Person HH 15.6% Five Person HH 13.3% Six or More Persons HH 22.4% Other HH size structure Households with a single adult and at least one child 13.0% Married couple household 70.2% Households with an equal proportion of adult male to adult female 38.2% Predominantly female households 35.1% Predominantly male households 26.8% Highest Education Less than middle school graduate 18.8% Middle school graduate/BECE 30.9% Some High School 2.7% Vocational Training 2.3% SSCE certificate/A level certificate 13.4% Training College/Polytechnic 4.3% Bachelors 3.7% Unknown 23.9% Household Monthly Income Income less than 200.01 ghc 43.4% Income between 200.01 ghc and 600 ghc 31.9% Income between 600.01 ghc and 1000 ghc 10.5% Income between 1000.01 ghc and 1200 ghc 2.9% Income greater than 1200 ghc 4.3% Unknown 6.9% Residential Location HH in urban area 42.7% HH in rural area 57.3% HH lives in city center 5.1% HH lives in town 35.9% HH lives in the suburb 32.0% HH lives along a major road 8.6% HH lives on the farm 6.8% HH lives near trans terminal 0.2% Other Location 8.1% Unknown 3.2 Western Region 9.6% Central Region 8.5% Greater Accra Region 14.3% Volta Region 9.5% Eastern Region 11 .4% Ashanti Region 18.8%
42 Table 2 1. Continued Socio Economic Variables Percent Frequency Brong Ahafo Region 9.8% Northern Region 10.4% Upper East Region 4.8% Upper West Region 3.0% Vehicle Ownership HH with Zero vehicle 64.5% HH owns at lea st a car 4.2% HH owns at least a bicycle 23.7% HH owns at least a motorcycle 7.2%
43 Figure 2 3 Frequency Distribution of the Total Number of Trips in the Last Seven Days per Mode Table 2 2 Male and Female Proportion of Trips in the Last Seven Days per Mode Trip Mode Male Female All adults No Trip Trip No Trip Trip No Trip Trip Bus 63% 37% 63% 37% 63% 37% Taxi 61% 39% 61% 39% 61% 39% Car 94% 6% 96% 4% 95% 5% Moto 89% 11% 96% 4% 93% 7% Bicycle 82% 18% 95% 5% 89% 11% Other 97% 3% 98% 2% 97% 3% 0 2000 4000 6000 8000 10000 12000 14000 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 28 30 31 32 34 35 36 38 40 Frequency Number of trips Bus Taxi Car Moto Bicycle Other
44 Figure 2 4 Frequency Distribution of the Total Number of Foot Trips on a Normal Day Figure 2 5 M ale and Female Proportion of Foot Trips on a Normal Day 0 500 1000 1500 2000 2500 3000 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 28 30 31 32 34 35 36 38 40 Frequency Number of trips 3% 97% 4% 96% 4% 96% No Trip Trip No Trip Trip No Trip Trip Male Female All adults
45 Figure 2 6 Proportion of Figure 2 7 Proportion of Non y Metropolitan Location
46 Table 2 3 Characteristics of Working Adults in Ghana Variable Proportion of households by trip mode All modes Car Bicycle Motorcycle Sample 9648 467 2748 840 Gender Male 46.80 50.70 51.00 53.10 Female 53 .20 49.30 49.00 46.90 Age Categories 18 24 13.90 14.10 14.40 13.60 25 44 56.10 60.40 55.90 62.70 45 64 25.00 24.00 24.70 20.70 >=65 5.00 1.50 4.90 3.00 Highest Education Level Less than middle school 19.10 8.40 15,5 13. 10 Middle school graduate/ BECE 26.00 23.30 15.80 19.20 Some High School 2.00 2.80 1.50 2.00 Vocational Training 2.00 3.60 1.20 1.70 SSCE certificate/ A level certificate 8.60 14.80 6.60 10.50 Training College/Polytechnic 2.70 7.90 1.70 4.30 Bache lors 2.50 18.60 1.30 4.00 Other 0.70 2.80 0,5 1.10 HH member currently in school 5.20 13.30 5.90 5.60 Employment Status Employee 1.30 0.90 1.20 1.50 Employer 0.00 0.00 0,5 0.10 Self Employed 52.20 53.30 56.80 55.40 Unpaid Family Work er 0.10 0.00 0.10 0.00 Apprentice 45.80 45.00 41.30 42.00 Other Employment Status 0.50 0.90 0.50 1.00 Household Monthly Income Income less than 200.01 ghc 38.10 5.40 43.30 32.00 Income between 200.01 ghc and 600 ghc 34.30 26.30 33.80 36 .40 Income between 600.01 ghc and 1000 ghc 12.20 19.70 10.30 15.20 Income between 1000.01 ghc and 1200 ghc 3.60 10.50 3.30 6.20 Income greater than 1200 ghc 6.10 32.50 6.50 7.40
47 Table 2 3 Continued Variable Proportion of households by trip mode All modes Car Bicycle Motorcycle Sample 9648 467 2748 840 Marital Status Married 68.40 67.70 76.70 81.00 Divorced/Separated 6.00 3.20 2.70 2.00 Widowed 5.20 2.40 3.50 1.70 Never married 20.40 26.80 17.10 15.40 Ethnicity Akan 47.00 67.70 23.00 25.70 Ga/Dangme 6.40 7.30 2.50 1.90 Ewe 13.20 12.60 9.00 8.20 Mole Dagbani 19.80 8.10 39.30 36.80 Other ethnicity 12.70 3.00 25.40 26.20 Religion Christian 72.10 88.70 52.10 53.30 Islam 17.70 7.90 32.80 37.30 Traditional 5.20 0.20 11.20 6.00 Other religion 5.10 3.20 3.90 3.50
48 Table 2 3 Continued Variable Proportion of households by trip mode All modes Car Bicycle Motorcycle Sample 9648 467 2748 840 Residential locat ion Household member in urban area 41.70 77.10 30.50 48.50 Household member in rural area 58.30 22.90 69.50 51.50 Household lives in city center 4.70 8.10 2.30 1.90 Household lives in town 36.00 28.10 27.70 35.80 Household lives in a subur ban area 32.40 53.50 26.70 28.60 Household lives along a major road 8.40 6.40 13.10 10.50 Household lives on the farm 8.40 1.30 14.80 9.80 Household lives near transport terminal 0.20 0.00 0.00 0.00 Other household location 8.60 1.90 14.60 12.70 Regional census Western 10.00 4.30 3.80 3.80 Central 6.30 4.30 1.20 1.20 Greater Accra 12.80 32.10 6.30 6.10 Volta 8.10 2.10 8.40 9.00 Eastern 11.00 13.90 4.10 3.30 Ashanti 20.00 28.90 9.70 10.50 Brong Ahafo 10.80 7.30 16.20 13.50 North ern 11.30 5.80 27.00 31.40 Upper east 6.60 0.40 15.90 11.80 Upper west 3.10 0.90 7.50 9.40 Sekondi Takoradi 1.40 1.10 0.10 0.20 Gomoa East (Central) 0.60 1.30 0.00 0.00 Accra Metropolitan Area 4.80 13.70 1.60 2.50 Ho Municipal 0.20 0.00 0.20 0.20 K waku North (Eastern Region) 0.80 0.40 1.30 1.50 Kumasi Metropolitan Area (Ashanti Region) 7.00 20.80 1.50 5.10 Techiman (Brong Ahafo Region) 0.90 0.90 1.20 1.40 Tamale Metro 1.70 4.50 4.00 9.80 Bolgatanga Municipal (Upper East Region) 0.90 0.00 1.80 1 .40 Wa Municipal (Upper West Region) 0.30 0.90 0.50 1.40
49 Table 2 3 Continued Variable Proportion of households by trip mode All modes Car Bicycle Motorcycle Sample 9648 467 2748 840 Vehicle Ownership Bicycle 28.48 17.10 100.00 57.60 Motorcycle 8.71 10.28 17.61 100.00 Car 4.84 100.0 0 2.91 5.71 Bus 0.44 3.00 0.18 0.80 Truck 0.39 1.93 0.55 1.55 Perception Variables Bus route convenient Yes 47.20 46.50 44.60 48.00 No 39.30 42.80 42.80 42.50 Satis fied with bus condition Yes 70.40 75.20 63.80 67.00 No 16.10 14..1 23.70 23.60 Satisfied with bus schedule frequency Yes 52.80 67.50 46.70 56.30 No 32.60 21.60 39.60 32.50 First important transport problems Transport fares too high 19.50 10.90 26.60 22.60 Long delays at transport station 21.70 14.60 19.90 18.50 Bad roads 38.80 34.50 39.00 38.80 Traffic jam 11.10 30.20 5.20 8.00 Other transport problems 3.10 4.90 4.30 5.80
50 Table 2 4 Characteristics of Non Working Adults in Ghana Variable Proportion of households by trip mode All modes Car Bicycle Motorcycle Sample 2849 148 763 229 Gender Male 46.60 60.10 55.00 51.50 Female 53.40 39.90 45.00 48.50 Age Categories 18 24 56.20 39.20 66.70 65.90 25 44 9.00 4.70 12.10 13.10 45 64 13.20 30.40 9.60 9.60 >=65 21.60 25.70 11.70 11.40 Highest Education Level Less than middle school 15.00 6.80 14.20 12.70 Middle school graduate/ BECE 20.20 20.30 12.70 10.50 Some High School 1.50 2.00 0.70 0.40 Vocational Training 1.80 6.80 1.20 1.30 SSCE certificate/ A level certificate 9.00 13.50 5.80 8.70 Training College/Polytechnic 2.60 10.80 2.20 3.90 Bachelors 2.10 13,5 1.20 1.70 Other 0.40 0.70 0.10 0.90 HH member currently in school 22.70 23.60 31.60 31.40 Household Monthly Income Income less than 200.01 ghc 39.40 6.10 40.10 28.40 Income between 200.01 ghc and 600 ghc 30.40 20.30 34.30 37.10 Income between 600.01 ghc and 1000 ghc 12.10 23. 00 9.40 14.00 Income between 1000.01 ghc and 1200 ghc 4.10 15.50 3.80 7.00 Income greater than 1200 ghc 6.90 26.40 8.50 10.00
51 Table 2 4 Continued Variable Proportion of households by trip mode All modes Car Bicycle Motorcycle Sample 2849 148 763 229 Marital Status Married 35.10 48.60 35.30 36.20 Divorced/Separated 4.40 2.70 2.00 2.20 Widowed 12.70 7.40 6.30 6.60 Never married 47.80 41.20 56.50 55.00 Ethnicity Akan 43.90 60.80 20.40 22.70 Ga/Dangme 8 .30 5.40 4.20 5.20 Ewe 14.20 14.20 8.70 7.40 Mole Dagbani 19.50 4.10 38.90 36.70 Other ethnicities 13.60 5.40 26.70 27.50 Religion Christian 72.60 95.30 52.70 49.30 Islam 17.90 2.00 33.20 40.60 Traditional 5.20 0.70 11.00 6.10 Ot her religion 4.40 2.00 3.10 3.90
52 Table 2 4 Continued Variable Proportion of households by trip mode All modes Car Bicycle Motorcycle Sample 2849 148 763 229 Residential location Household member in urban area 45.10 83.80 33.90 50.70 Household member in rural area 54.90 16.20 66.10 49.30 Household lives in city center 5.50 10.80 2.40 0.90 Household lives in town 34.90 22.30 25.60 28.40 Household lives in a suburban area 31.10 55.40 26.90 31.40 Household lives along a major road 9.80 10.10 14.70 14.00 Household lives on the farm 8.00 0.00 13.80 10.90 Household lives near tr ansport terminal 0.20 0.00 0.30 0.00 Other household location 8.80 0.70 15.70 13.50 Regional census Western 8.20 0.70 2.50 2.20 Central 6.50 1.40 0.90 0.40 Greater Accra 14.40 41.90 8.30 8.70 Volta 9.70 0.70 8.80 7.90 Eastern 11.90 2 5.00 5.40 2.60 Ashanti 19.30 20.90 8.30 11.80 Brong Ahafo 9.00 6.10 14.20 15.70 Northern 11.50 0.70 28.70 30.60 Upper East 6.70 1.40 17.00 10.00 Upper West 2.90 1.40 6.00 10.00 Sekondi Takoradi 1.40 0.00 0.10 0.00 Gomoa East (Central) 0.60 0.00 0.0 0 0.00 Accra Metropolitan Area 6.20 16.20 1.60 4.40 Ho Municipal 0.20 0.00 0.10 0.00 Kwaku North (Eastern Region) 0.50 0.00 0.90 0.90 Kumasi Metropolitan Area (Ashanti Region) 6.70 20.30 1.00 6.10 Techiman (Brong Ahafo Region) 0.60 0.00 1.20 1.30 Tam ale Metro 2.10 0.70 5.40 10.90 Bolgatanga Municipal (Upper East Region) 1.20 0.70 2.60 1.30 Wa Municipal (Upper West Region) 0.20 1.40 0.10 1.70
53 Table 2 4 Continued Variable Proportion of households by trip mode All modes Car Bicycle Motorcycl e Sample 2849 148 763 229 Vehicle Ownership Bicycle 26.78 17.57 100.00 56.64 Motorcycle 8.04 7.43 17.30 100.00 Car 5.19 100.00 3.41 4.80 Bus 0.56 2.70 0.50 3.93 Truck 0.49 2.03 0.70 2.62 Perception Variables Bus rou te convenient Yes 45.30 38.50 40.90 43.70 No 34.90 41.20 37.90 37.10 Satisfied with bus condition Yes 64.90 64.90 55.00 56.80 No 15.50 15.50 24.20 24.90 Satisfied with bus schedule frequency Yes 49.70 59.50 41.00 47.60 N o 29.30 20.30 36.70 31.00 First important transport problems Transport fares too high 21.00 16.90 29.10 28.80 Long delays at transport station 21.20 12.20 18.90 21.40 Bad roads 37.30 34.50 40.10 32.30 Traffic jam 11.40 25.70 4.50 8.30 Ot her transport problems 3.30 9.50 2.80 5.20
54 Table 2 5 Trip Generation Model for Working Adults in Ghana (Vehicles Owned) Variables Car Bicycle Motorcycle Coeff T stat Coeff T stat Coeff T stat Constant 1.213 2.300 2.621 4.680 5.100 12.7 20 Socio Economic Variables Household size 0.140 3.260 0.101 2.280 Total number of adults in household 0.199 5.990 0.516 5.680 Female 0.785 4.150 1.731 16.840 2.129 12.930 Age in years 0.044 4.060 0.015 3.450 0.024 3.230 Household member currently attending school 0.954 3.820 Less than middle school graduate 0.374 2.640 0.460 2.110 Middle school graduate/BECE 0.845 5.940 Some High School 0.886 2.090 SSCE certificate/A level certificate 0.674 2.610 0.677 3.280 0.582 2.400 Training College/Polytechnic 1.186 3.420 1.241 3.310 0.806 2.300 1.045 4.190 2.362 5.060 Income less than 200.01 ghc 2.226 4.390 0.304 3.050 Income between 200.01 g hc and 600 ghc 1.053 4.810 0.403 2.550 Income between 1000.01 ghc and 1200 ghc 0.911 2.940 Never married 0.759 2.860 Married 0.552 2.200 0.461 2.220 0.658 2.940 Divorced/Separated 1.206 2.050 Vehicle Ownership Car owned by household 0.622 3.230 0.697 3.020 Bus owned by household 2.922 2.980 Truck owned by Household 2.985 2.180 Residential Location Household lives in city center 1.973 4.900 Household lives on farm 0.842 3.210 Accra Metropolitan Area (Greater Accra Region) 1.533 5.130 1.097 2.600 Kumasi Metropolitan Area (Ashanti Region) 1.018 2.430 Tamale Metro (Northern Region) 1.044 4.090 Wa Municipal (Upper West Region) 1.800 2.590 Time to walk to the nearest taxi rank (minutes) 0.051 2.610 Time to walk to the nearest bus stop (minutes) 0.206 2.700 0.042 1.960 Sample Size 467 2748 840 LL(B) 1001.712 4985.698 1651.129 LL( c ) 1074.43 5160.922 1761.364 Pseudo R square 0.067 0.034 0.0626
55 Table 2 6 Trip Generation Model for Working Adults in Ghana ( Foot, Taxi, Bus) Variables Foot Taxi Bus Coeff T stat Coeff T stat Coeff T stat Constant 1.886 21.800 1.792 7.400 0.918 12.030 Socio Economic Variables Household size 0.018 5.060 0.054 6.430 Total number of employees/workers in household 0.080 4.480 Female 0.098 2.260 Age in years 0.001 6.940 Household member currently attending school 0.194 4.390 0.377 3.440 Less than middle school graduate 0.381 6.080 0.276 4.270 Middle school graduate/BECE 0.075 3.400 0.590 10.000 0.498 8.350 Some High School 0.402 2.640 0.391 2.420 Vocational Tra ining 0.162 2.550 0.939 6.460 0.516 3.300 SSCE certificate/A level certificate 0.068 2.000 0.666 7.940 0.663 7.730 Training College/Polytechnic 0.695 5.200 0.544 3.820 0.131 2.220 0.820 5.880 0.453 3.060 Self Employed 0.072 4.010 Income less than 200.01 ghc 0.077 3.820 0.542 9.360 0.294 6.030 Income between 200.01 ghc and 600 ghc 0.165 3.030 Income between 600.01 ghc and 1000 ghc 0.071 2.460 Income between 1000.01 ghc and 1200 ghc 0.138 2 .800 Never married 0.258 4.540 Vehicle Ownership Car owned by household 0.141 2.240 Motorcycles owned by household 0.044 2.070 0.224 3.700 0.121 1.960 Bicycles Owned by household 0.024 2.020 0.174 9 .090 0.095 3.620 Bus owned by household 0.424 5.100 0.523 2.710 Truck owned by Hhousehold 0.609 2.080 Residential Location Household lives in city center 0.516 11.920 0.360 2.760 Household lives in town 0.670 7.360 Household lives in a suburban area 0.258 12.230 0.777 8.320 0.186 3.490 Household lives along a major road 0.099 2.990 0.565 5.160 0.189 2.260 Household lives on farm 0.256 7.340 0.273 2.360 0.348 3.790 Sekondi Takoradi 0.43 5 2.530 0.576 3.140 Accra Metropolitan Area (Greater Accra Region) 0.346 7.560 0.863 7.740 Ho Municipal (Volta Region) 1.321 2.530 Kwahu North (Eastern Region) 0.897 8.120 0.955 3.440 Gomoa East (Central Region) 0.443 3.420 Kumasi Metropolitan Area (Ashanti Region) 0.253 2.950 0.701 7.520 Techiman (Brong Ahafo Region) 0.722 2.950 Tamale Metro (Northern Region) 1.128 7.030 0.923 4.750 Bolgatanga Municipal (Upper East Region) 0.862 2.940 Wa Municipal (Upper West Region) 1.453 2.680 0.973 2.340 Time to walk to the nearest taxi rank (minutes) 0.026 6.960 0.149 16.150
56 Table 2 6 Continued Variables Foot Taxi Bus Coeff T stat Coeff T stat Coeff T stat Time taken to walk to the nearest station/boarding point 0.007 3.570 Sample Size 9648 9648 9648 LL(B) 28331.415 15933.337 15737.146 LL( c ) 28720.217 16589.564 16146.289 Pseudo R square 0.0135 0.0396 0.0253
57 Table 2 7 Trip Generation Model for Non Working Adults in Ghana (Vehicles Owned) Variables Car Bicycle Motorcycle Coeff T stat Coeff T stat Coeff T stat Constant 2.98 2.47 4.10 7.56 6.06 2.66 Socio Economic Variables Total number of ad ults in household 0.22 3.67 Female 1.42 7.00 1.99 3.65 Age in years 0.04 2.39 0.05 7.48 Household member currently attending school 0.88 3.36 2.07 3.43 Less than middle school graduate 0.65 2.22 Middle school g raduate/BECE 1.61 4.94 SSCE certificate/A level certificate 1.45 3.22 Training College/Polytechnic 1.01 2.37 4.03 3.02 1.12 2.86 Income less than 200 ghc 2.39 3.20 Income between 200.01 ghc and 600 ghc 1.94 2.72 Income between 600.01 ghc and 1000 ghc 2.18 2.75 Never married 3.46 4.31 2.87 2.26 Married 3.16 2.60 Vehicle Ownership Bicycle Owned 0.51 2.23 Bus owned by household 0.57 2 .40 Residential Location Household lives in city center 1.29 2.03 Household lives in town 3.96 2.42 Household lives in a suburban area 3.81 2.85 1.84 3.19 Accra Metropolitan Area (Greater Accra Region ) 1.50 4.28 2.15 2.29 Kumasi Metropolitan Area (Ashanti Region) 2.82 1.95 Time to walk to the nearest taxi rank (minutes) 0.21 2.19 Time to walk to the nearest bus stop (minutes) 0.37 3.30 Sample Size 148 763 229 LL(B) 328.612 1370.465 309.743 LL( c ) 361.715 1416.507 339.765 Pseudo R square 0.0915 0.0325 0.0884
58 Table 2 8. Trip Generation Model for Non Working Adults in Ghana (Foot, Taxi, Bus) Variables Foot Taxi Bus Coeff T stat Coeff T stat Coeff T stat C onstant 2.34 22.70 1.19 3.68 0.63 2.84 Socio Economic Variables Household size 0.02 3.83 0.07 4.09 0.06 3.15 Female 0.22 2.05 Age in years 0.01 14.12 Household member currently in school 0.15 3.26 0.50 3.51 Less than middle school graduate 0.53 3.54 0.50 3.27 Middle school graduate/BECE 0.52 3.75 0.67 5.00 Some High School 0.86 2.13 Vocational Training 1.00 2.69 SSCE certificate/A level certificate 0.64 3.62 0.56 3.03 Trai ning College/Polytechnic 0.65 2.12 0.67 2.06 Income between 200.01 ghc and 600 ghc 0.33 2.76 Income between 600.01 ghc and 1000 ghc 0.11 2.10 0.39 2.49 Income between 1000.01 ghc and 1200 ghc 0.88 3.6 4 Income greater than 1200 ghc 0.66 3.28 Married 0.27 2.47 Vehicle Ownership Motorcycle Owned 0.11 2.19 Bicycle Owned 0.13 2.52 Bus owned by household 1.73 2.90 0.99 2.26 Resi dential Location Household lives in city center 0.46 5.26 0.67 2.57 Household lives in town 0.12 2.26 0.89 5.20 Household lives in a suburban area 0.13 2.53 1.01 5.71 0.30 2.44 Household lives along a major road 0.70 3.39 Household lives on farm 0.37 4.93 Sekondi Takoradi 0.36 2.38 0.81 2.05 1.03 2.53 Accra Metropolitan Area 0.25 3.15 0.58 2.70 1.06 4.66 Kwahu North 1.30 4.64 Kumasi Metropolitan Area 0.81 4.05 1.08 5.09 Tamale Metro 1 .10 3.43 2.48 4.63 Minutes to walk to the nearest taxi rank 0.02 2.76 0.16 6.97 0.10 2.26 Minutes to walk to the nearest bus stop 0.02 2.37 0.05 2.10 Sample Size 2849 2849 2849 LL(B) 8024.228 3637.324 3445.976 LL( c ) 8198.831 38 06.916 3554.867 Pseudo R square 0.0213 0.0445 0.0306
59 CHAPTER 3 VEHICLE OWNERSHIP MODEL Literature Review This section provides a summary of previous studies on vehicle ownership which describes the household vehicle fleet composition in developing co untries. Vehicle ownership models are broadly explored in the United States and other developed countries. There is a vast body of literature on modeling household vehicle ownership (see for instance Train and Lohrer, 1982; Potoglou and Susilo, 2008 ) in t he United States. However, there is little research on vehicle ownership n in developing countries like Ghana, as Ghana mostly researches on air quality, safety, traffic congestion, and other transport challenges. Therefore, we present a detail discussion of these studies focusing on Ghana and other developing countries. This section also focuses on the literature on vehicle ownership models considering the methodology often used and some useful findings on the main empirical factors. Vehicle ownership mode ls can be developed both at the aggregate and disaggregate level s In developed countries, they are often modeled using household disaggregate data and by employing discrete choice models. Disaggregate models are characterized by ordinal categorical respo nse variables which are ordered choice models and nominal response variables which are unordered choice models. The ordered response model assumes the household vehicle ownership as a continuous latent variable and the most used modeling methodology or th e ordered response structures is the ordered probit model. The unordered response models employ a random utility maximization where households chose the ownership level that has the maximum utility or satisfaction (Potoglou and Susilo 2008 ). The most used modeling methodology for the household
60 vehicle fleet composition is the multinomial logit model (MNL) for these unordered structures. Furthermore, most developed countries like the USA and UK develop their vehicle ownership models using data from their na tional household travel surveys; however, most developing countries do not have periodical travel surveys that are conducted on the national levels largely due to the expense in data collection and the lack of research interests. From studies, such as Ben Akiva et al. (1981), car ownership models are often modeled based on the socio economic characteristics of the household, and travel related characteristics. In understanding the factors that influence vehicle ownership in low income countries, Button et a l. (1993) established that vehicle ownership generally increases as low income countries become richer or more stable economically. Using the 2005 Dutch National Travel Survey (NTS) data set in the study by Potoglou and Susilo (2008) on the comparison of vehicle ownership models, the number of workers, household income, life cycle classification of the household, and location positively influenced the household vehicle ownership. Households with more workers are more likely to own more cars, likewise high e r income households. Households with a couple and children compared with single households are more likely to own two cars than a single or greater than three vehicles. Also, extended family households are more likely to own multiple cars than single hous eholds. Further, households in non urbanized locations are more likely to own multiple vehicles than households in highly urbanized areas.
61 Further studies by Schimek (1996) and Matas et al. (2009) have also looked at how land use characteristics and spatia l variables influence vehicle ownership and travel behavior. In the US, it is observed that people in highly dense areas travel mostly by foot or public transit Schimek (1996). Further, they modeled the vehicle ownership and used that as an explanatory var iable for their vehicle use model. Household characteristics together with neighborhood characteristics like density and other travel characteristics were the main factors considered for their study. From their study, households in highly dense areas own j ust fewer vehicles thus make fewer trips than households in low density areas. Households in the city center own about 16% fewer vehicles than households in the suburbs. In addition to household income, household size and the number of workers in a househo ld being significant factors of the household vehicle ownership, households with the head being less than 35 years own fewer cars than older household heads. Accessibility variables have also been considered in modeling vehicle ownership. Bhat and Guo (200 7) capture the effect of accessibility variables like transit, work, shopping and recreation on the vehicle ownership model formulation. In Spain, job accessibility used as an explanatory variable in their vehicle ownership model demonstrated through a sim ulation exercise to capture its influence in the vehicle ownership model show s that vehicle ownership decreases as accessibility becomes better. (Matas et al. 2009 ) Their findings also show that households with more adult members especially workers are mo re likely to own at least a car. Another interesting finding is that households with the head of the family being a male, married, an employer or employed in a managerial occupation are more likely to own at least one
62 car compared to an unskilled working h ousehold head. Kumar and Krishna Rao (2006) studied the effects of projected household income, household size, travel time, travel cost, total cost of a car including maintenance cost through a stated preference experiment of car ownership that was conduct ed in Mumbai Metropolitan Region. They also tested the validity of the SP model by using the generated model to predict the observed choice behavior and established that there were no significant differences between the predicted and observed choices. Ther efore, the SP modeling approach is viable to employ in dev eloped countries where there may not be actual observed data. Data Data Structure The data used in this study come s from the 2012 Ghana Transport Indicator Database Survey which was conducted by th e Ghana Statistical Service. This data which represents the second phase of the first ever nationwide household based transport survey was conducted over a 3 month period between September 2012 and December 2012. The survey was administered to 6000 househo lds which represent all the households in Ghana from all the ten regions. The data includes some basic socio economic characteristics of the household (such as size and income) in addition to variables describing the location of the household such as regio n and district where the household lives. The focus of this analysis is on household vehicle ownership. The survey which explicitly collected data on the transport behaviors of Ghanaians also included data on the various vehicle fleet that was owned by the household. The database was based on the data of the individual household members. The dataset provides information for 23238 individuals from 6000 households selected from a total of 400 Enumeration
63 Areas (EA). Since the dataset was based on the personal level, the person level dataset was aggregated to the household level (taking into account the vehicle ownership of each household) based on a unique HHID that was generated. The data doesn't explicitly provide any variable that identifies each household member belonging to the same household. Therefore, with supplemental information on the enumeration areas listing that was used for the survey, which was provided by the Ghana Statistical services, a unique Household ID code was generated to match the hous ehold members with their respective households. The data provides a variable as household number (HHID); however, a frequency distribution of this variable showed that there were 20 household number s with 15 of them each having more than 1450 households. F rom a further probe into this variable, we were able to identify that the household numbers referred to the number (starting from 1) assigned to each household that was randomly selected in each enumeration area for the interview. Therefore, a unique code combining each HHID in their respective EA was generated. For example, a code of 2312 refers to an EA of 231 and an HHID of 2). With this code generated, household members were linked to their respective households and we generated some household character istics such as household size, and the total number of adults, children, workers and students in a household. Data Descriptives There were 6000 households, but the dataset has a sample of 5921 after a thorough cleaning. Table 3 1 provides a full descriptiv e statistics of the explanatory variables considered for analysis that describe the vehicle ownership of all households. From the data, other variables relating to the household structure were derived to understand the influence of these variab les of the vehicle fleet
64 composition of a household. The interesting new variables derived for this analysis include single pare nt household and married couple households. Also, to capture the role of gender in vehicle ownership, "predominantly male househ olds," "predominantly female households" and households with an equal proportion of adult male to adult female" variables were generated. On average, there are 3.99 members per household with 2.05 workers per household on average. The workers comprise of e mployees, employers, self employed workers, unpaid family worker, and apprentices. About 68.2% of the households have children with an average of 2.67 children per household. With respect to gender, 3 5.1 % of the households' adults are predominantly women, 26.8% are predominately men with the remaining representing households with an equal proportion of male and female adults. In terms of location, 42. 7 % of the households live in urban areas whereas 57. 3 % live in rural areas. The topmost 5 regions capturing the households' locations are Ashanti Region (18. 8 %), Greater Accra Region (14 .3 %). Northern Region (1 0 .4%), Eastern Region (11. 4 %) and Brong Ahafo Region (9. 8 %). Almost half of the households are low income households that earn less than 200ghc ( $104 per 2012 cedi to dollar conversion rate) per month. The vehicle fleet from Ghana shown in the data are car, bicycle, motorcycle, bus, truck and others, however, for this analysis purpose, only car, bicycle, and motorcycle is considered as less than 1% of the h ouseholds own at least a bus or truck and thus may show little impact in this analysis. Therefore, the choice set considered for model ing is presented in Figure 3 1 which also shows the proportion of these choices on a regional level.
65 Modeling Methodology This section describes the structure and assumptions of the methodology employed for the vehicle fleet ownership decision for all households in Ghana. The multinomial logit (MNL) modeling approach for the vehicle ownership decision modeling is used. The MNL model is the most common unordered response discrete choice model based on the random utilization principle. That is, households assign a utility function to the choice alternatives (zero vehicles, car, motorcycle, single bicycle, multiple bicycles, m ixed fleet) and choose the alternative with the maximum utility. The utility function is a linear function of the household characteristics such as household size structure, income, and residential location. Additionally, the random utility maximization pr inciple also incorporates the effects of unobserved factors not considered in the utility function. The random utility maximization framework is represented below: Where = total utility of household n for choos ing vehicle i = deterministic or observed component representing the parameter estimate and = random or unobserved component of utility for household n for choosing vehicle i From this utility function, the choice probability, tha t is the probability of a household k choosing vehicle i is derived: The MNL model specification is based on the assumption that there is no correlation or similarities between the choice alte rnative. This assumption, known as the
66 Independence from Irrelevant Alternatives (IIA) property assumes that for a specific individual, the ratio of the choice probabilities of any two alternatives is a function of only the characteristics of those two alt ernatives and is independent of all other alternatives present in the choice set. The MNL model was developed to determine the factors that influence a developed using the S PSS statistical software with the six choice alternatives and the choice alternative of zero vehicles kept as the reference variable. All the household variables detailed in Table 3 1 were considered as explanatory variables for the model but the model dev eloped as shown in Table 3 2 only highlights the explanatory variables showing at least a 95% confidence Results The results of the vehicle ownership models developed in this section seek to understand the influence of household characteristics on their v ehicle ownership. This study estimated the household vehicle ownership levels across all the ten regions in Ghana and we present the results of the MNL model with the coefficients of the estimate and their t statistics in Table 3 2. From the results, the most consistent significant variable across all vehicle ownership levels were the regional location variables. The results show that households owning at least a car show a very strong positive correlation with the regional location variables. However, Gr eater Accra, Ashanti, and Eastern regions are the strongest predictors. These three regions are the wealthiest in the country as reported by the UNICEF, (Cooke et al., 2016) therefore, this result was expected. This further endorses income as a strong sign ificant variable in impacting car ownership.
67 The negative coefficient on the income variables relative to the highest income variable shows that low in income households are less likely to own a car. Furthermore, lowest income households (households with a monthly income less than 200 ghc) are the least likely to own a car. As household incomes increase their propensity to own a vehicle increases as well. On the other hand, all the other household ownership levels show a negative association with the regi onal location variables. This shows that, generally, Ghanaians have a lesser tendency to own at least a two wheeler (bicycle or motorcycle) as they pride in owning a personal car instead. Therefore, households living along a major road are the ones more l ikely to own a motorcycle or multiple bicycles. Additionally, households in the poorest regions of Ghana, that is, the northern part of the country (Northern, Upper East, and Upper West regions) are most likely to own a bicycle, motorcycle or a mixed fleet The two wheeler is the predominantly used mode in the northern part of the country meanwhile in the wealthier regions like Greater Accra where car ownership is high, two wheelers ownership is lower due to safety reasons as well. First of all, the road ne tworks not designed with cyclists in mind together with the aggressiveness of drivers in these wealthier regions generally deter those households from owning a two wheeler as the two wheelers will have to compete with other road users like trucks, buses, c ars, and even road side hawkers. In terms of education, generally, households with higher levels of education level are more likely to own at least a personal car and a motorcycle except for households with the highest education level being Training Coll ege/Polytechnic. Such households are less likely to own a car compared with households with the highest education of
68 SSCE certificate/ A Level certificate although training college degrees are higher than SSCE certificates. The household structure signif icantly affects the single and multiple bicycle ownership only. Smaller households are less likely to own multiple vehicles compared with larger households. All this being equal, the results show that married couple households are more likely to own a mot orcycle than unmarried couple households. Additionally, households with predominantly female adults are less likely to own at least a two wheeler compared to households with an equal proportion of male to female adults. Additionally, predominantly male hou seholds are more likely to own a motorcycle. This finding is however expected as females are usually more concerned about safety and appearance in traffic thus less likely to ride bicycles. Possible reasons for the decline of vehicle ownership for predomin antly female households could be the social stigma attached to women riding bicycles. In Ghana, there is a socio cultural myth that females who ride a bicycle are prone to lose their virginity, such females are there is a social stigma attached to a female riding a bicycle as in the past, cultural practices identify bicycle ridership to men. Findings from Calvo (1994) and Peters (2001) also support these explanations. It is, however, useful to note that these so cial myths and other cultural norms have evolved over the years and there is a possibility of an increase in bicycle ownership in predominantly female adult households. The effect of the total number of children and workers in a household are is anticipat ed. Dependent on the location of the household, households with more adult workers are more likely to own multiple bicycles to serve the many workers. Also, the
69 larger the number of children in a household, the more likely it is for that household to own a mixed fleet of vehicles. Summary and Conclusion Vehicle ownership is a key determinant in the study of travel demand. With the growing needs of vehicle ownership in Ghana, it is important to accurately model the factors that influence the ownership of the different vehicle classes. This is also essential in making future predictions for traffic congestion, especially since traffic congestion is one of the main transport issues in the urbanized areas of Ghana. This study employs the multinomial logit model (MNL) to model the vehicle ownership choices (car, motorcycle, single bicycle, multiple bicycle, mixed fleet). The model considered the commonly used socio economic variables of the household such as household size, household income, and residential locati on. In addition to these frequently used variables, this model also considers the highest education of adults in the household, the gender proportion of households, and if a couple lived in the household. The estimated model results confirm and support the identified findings from several publications. The effect of income and education levels of a household significantly influencing household vehicle ownership are established findings. Some other notable results from the model also show that the effect of the household residential location significantly impacts the kind of vehicle a household is likely to own with the exception of households that own at least one car. It was interesting to find out that households with predominantly female adults are more likely to own at least a two wheeler compared to households with an equal proportion of male to female adults. Also, married couple households are more likely to own a motorcycle than unmarried
70 couple households. These findings help us understand clearly h ow household structure influences vehicle ownership and consecutively account for these in developing transportation plans.
71 F igure 3 1 Proportion of Vehicle Fleet Ownership by Regional Location
72 Table 3 1 Characteristics of Households Socio Economic Variables Percent Frequency Household size structure Single Person HH 20.3% Two Person HH 13.3% Three Person HH 15.1% Four Person HH 15.6% Five Person HH 13.3% Six or More Persons HH 22.4% Other HH size structure Households with a single adult and at least one child 13.0% Married couple household 70.2% Households with an equal proportion of adult male to adult female 38.2% Predominantly female households 35.1% Predominan tly male households 26.8% Highest Education Less than middle school graduate 18.8% Middle school graduate/BECE 30.9% Some High School 2.7% Vocational Training 2.3% SSCE certificate/A level certificate 13.4% Training College/Polytechnic 4.3% B achelors 3.7% Unknown 23.9% Household Monthly Income Income less than 200.01 ghc 43.4% Income between 200.01 ghc and 600 ghc 31.9% Income between 600.01 ghc and 1000 ghc 10.5% Income between 1000.01 ghc and 1200 ghc 2.9% Income greater than 1200 ghc 4.3% Unknown 6.9% Residential Location HH in urban area 42.7% HH in rural area 57.3% HH lives in city center 5.1% HH lives in town 35.9% HH lives in the suburb 32.0% HH lives along a major road 8.6% HH lives on the farm 6.8% HH lives ne ar trans terminal 0.2% Other Location 8.1% Unknown 3.2 Western Region 9.6% Central Region 8.5% Greater Accra Region 14.3% Volta Region 9.5% Eastern Region 11.4% Ashanti Region 18.8 %
73 Table 3 1. Continued Socio Economic Variables Percent Fre quency Brong Ahafo Region 9.8% Northern Region 10.4% Upper East Region 4.8% Upper West Region 3.0% Vehicle Ownership HH with Zero vehicle 64.5% HH owns at least a car 4.2% HH owns at least a bicycle 23.7% HH owns at least a motorcycle 7.2%
74 Table 3 2 MNL Model on Vehicle Ownership Variables Car Motorcycle Single Bicycle Multiple Bicycles Mixed Fleet Coeff. Sig. Coeff. Sig. Coeff. Sig. Coeff. Sig. Coeff. Sig. Constant 24.059 0.000 0.073 0 .942 0.531 0.433 1.645 0.127 1.900 0.052 Household (HH) size HH size = 1 2.555 0.003 HH size = 2 1.109 0.050 HH size = 5 0.510 0.017 0.638 0.049 HH size = 6+ Ref Re f. Re f Ref Ref Ref Ref Ref Ref Ref Married couple household 0.927 0.001 Predominantly female households 0.891 0.002 0.717 0.000 0.683 0.009 0.903 0.002 Predominantly male households 0.635 0.009 Equal proportio n of male to female adult Ref. Ref. Ref. Ref. Ref. Ref. Ref. Ref. Ref. Ref. Total number of children 0.168 0.033 Total number of workers 0.258 0.004 Highest Education Less than middle sc hool graduate 2.111 0.000 2.093 0.000 Middle school graduate/BECE 1.583 0.000 1.193 0.000 Some High School 1.174 0.035 SSCE certificate/A level certificate 1.258 0.008 1.865 0.005 Training Coll ege/Polytechnic 1.400 0.000 0.976 0.004 0.894 0.026 Bachelor's Degree 0.989 0.005 Others Ref. Ref. Ref. Ref. Ref. Ref. Ref. Ref. Ref. Ref. Household Monthly Income Income less than 200.01 ghc 3.406 0.000 0.552 0.039 0.854 0.037 1.651 0.000 Income between 200.01 ghc and 600 ghc 2.106 0.000
75 Table 3 2. Continued Variables Car Motorcycle Single Bicycle Multiple Bicycles Mixed Fleet Coeff. Sig. Coeff. Sig. Coeff. Si g. Coeff. Sig. Coeff. Sig. Income between 600.01 ghc and 1000 ghc 1.367 0.000 Income greater than 1200 ghc Ref. Ref. Ref. Ref. Ref. Ref. Ref. Ref. Ref. Ref. Household Location HH lives along a major r oad 1.025 0.050 1.064 0.011 Others Ref. Ref. Ref. Ref. Ref. Ref. Ref. Ref. Ref. Ref. Regional Location Western Region 22.483 0.000 3.702 0.000 2.548 0.000 3.038 0.000 5.417 0.000 Central Region 23.620 0.000 3.885 0.000 3.361 0.000 6.341 Greater Accra Region 24.543 0.000 3.082 0.000 2.597 0.000 2.192 0.002 4.558 0.000 Volta Region 22.767 0.000 1.986 0.000 1.593 0.000 1.376 0.043 3.710 0.000 Eastern Region 23.983 0.000 4.025 0.000 2.884 0.000 3.534 0.000 4.830 0.000 Ashanti Region 24.039 0.000 3.229 0.000 2.702 0.000 2.973 0.000 5.186 0.000 Brong Ahafo Region 23.150 0.000 1.870 0.005 1.006 0.100 3.406 0.000 Northern Region 1.258 0.007 Upper Eas t Region 1.365 0.007 Upper West Region Ref. Ref. Ref. Ref. Ref. Ref. Ref. Ref. Ref. Ref. Number of observations 5921 Log likelihood, constant only 7417.898 log likelihood, final model 5706.772 Note: Ref. = Reference
76 CHAPTER 4 PUBLIC TRANSPORT USERS PERCEPTION Literature Review This section presents a summary of the literature on public transport (PT) users perception. There is a substantial body of literature on PT users perception of safety and security, integrated public transp ort and PT trip attributes like travel time on public transportation (see for instance, Perone and Tucker (2003), Van et al. (2010) Chowdhury Subeh, et al.(201 8 )) More specifically, the literature synthesis focuses on the key contributors influencing tr ansport users perception, the methodologies often employed in modeling perceptions and some useful findings on the main empirical factors. Transport users perceptions are generally difficult to measure as they are based on subjective opinions on differen t aspects of public transportation like transport quality, safety, transport availability, trip characteristics etc. These subjective opinions could socio economic characteristics and their current state of mind during the period their opinions were sought for Therefore, perceptions are not uniform due to the several contributing factors that may influence them. Several studies have looked at how the various aspects of public transport ation like bus transfers, delays, tra nsport fares, commute behaviors and perceptions. In understanding how perceptions of PT services influence PT users to use routes with transfer s Chowdury and Cedar (2013) developed a user preference survey to measure the PT satisfaction with transfer routes in addition to other questions that targeted their perceived ease of using a route with transfers. Also, non transfer route users were asked questions on which key trip attributes
77 improvements would increase their w illingness to use transfer routes. From their study, they identified the improvements to the fare and ticketing system, and operational trip characteristics to increase the willingness to use transfer routes. Kumar et al. (2011) addresses some safety perc eptions by studying the security perception of its PT users in multi modal trips. Through a summary descriptive of their data, PT users were most satisfied with the bus stop waiting security and least satisfied with the security at transfer from bus to met ro interchanges Furthermore, several studies by Currie et al. (2013), Loukaitou Sideris (2009 ) explain how personal safety perceptions influence the patronage of public transport. Morse and Benjamin (1996) Currie et al. (2013) and Loukaitou Sideris (2009 ) i dentif y gender to influence a PT Loukaitou Sideris (2009), young females have a higher perception of feeling unsafe on public transportation compared to young men. Carrel et al. (2013) d evelop e d an online survey for current and past users of the San Francisco Municipal Transit Authority (MUNI) to reduce PT use could be influenced by previous negative experience on the unreliability of the PT. From their ordinal probit mod el findings, the y i dentified key socio econo mic variables of age, gender and family status were insignificant in the model. Rather, current users of MUNI for less than one year had a higher propensity to continue using the MUNI. Furthermore, users who expe rienced long delays at transfer stops and delays due to operational issues by the transit agencies were more likely to reduce MUNI ridership An interesting f inding from their paper shows that occasional
78 inaccurate predictions t ime in formation do not reduce the users likelihood of using the transit service. Due to the growing need for improving public transportation in developing countries to help curb congestion and reduce the dependency on private cars Belwal and Belwal (2010) ex amine the perception of people on establishing an effective public transport system in Oman. An opinion survey was conducted on both PT users and non users to understand their perception differences as well. Their findings show that difference between PT users and non users in their satisfaction with PT experience. Also, a larger proportion of the respondents perceive PT as convenient and affordable although females consider PT services to be inconvenient. Studies on perceptions, esp ecially relating to PT are usually done by conducting some sort of opinion surveys to accurately capture perception differences and identify any inconsistencies. Studies by Chowdury and Cedar (2013), Kumar et al. (2011), Chowdhury et al. (2018 ) all conduct ed user preference surveys. Chowdury and Cedar (2013) use d a 5 point Lik ert scale (1=Very Poor, 2=Poor, 3=Neutral, 4=Good, and 5=Very Good) to measure their perceptions in term of satisfaction Likewise Kumar et al. (2011) also ranked user perceptions on a 5 point scale. In examining the perceptions of integrated public transport systems from the PT users and policymakers, Chowdhury et al. (2018) designed both a user preference survey for PT users and a semi structured interview for policymakers. For their s urvey design, the PT users perceptions wer e ranked on a scale of 1 9 ( Figure 4 1 ). The survey captured questions on the socio economic characteristics of the PT user in addition to trip attributes like average travel time and trip frequency al. (2010) d esign ed both a survey with questions on
79 and developed an ordered probit model to examine the overall quality of the transport service and identify the other factors that are likely to influence the perception of the overall service quality. The service reliability and waiting time were perceived to be some of the strongest factors of evaluating service quality, however, users perceptions about service quality were susceptible to change upon con sidering other strong variables. Generally, users perception on PT are largely influenced by several factors depend ing on the location, the route, users prior experience of PT, the transport service and conditions than the demographics of the users. How ever, this study aims to accurately capture the effect of the socio economic characteristics of the PT users their vehicle ownership levels and other trip attributes to understand the factors influencing the perceptions of PT. Data Data Structure This st udy used data from the 2012 Ghana Transport Indicator Database Survey which was conducted by the Ghana Statistical Service. This data which represents the second phase of the first ever nationwide household based transport survey was conducted over a 3 mon th period between September 2012 and December 2012. The survey was administered to 6000 households which represent all the households in Ghana from all the ten regions. The data includes some basic socio economic characteristics of the household (such as s ize and income) in addition to variables describing the location of the household such as region and district where the household lives. Perception variables with respect to public transportation usage are also included in the data.
80 However, for this analy sis, the focus is on understanding the perceptions of PT by public transportation users. The data highlights some perception questions which were asked to all respondents. However, some of the questions like reliability and main transport obstacle/difficul ty are specific to a trip purpose, but there is significant missing data in these variables. The significant missing information is largely due to the fact that the main transport mode for those purposes is the foot. For every household member that had ev er used a bus ( referred to as PT ) before, they were asked perceptions about the convenience of bus route, satisfaction of bus condition, satisfaction of transport availability and satisfaction of bus frequency schedule. Data Descriptives There were 6000 h ouseholds but the dataset has a sample of 5921 after a thorough cleaning. This analysis however focuses on the perceptions of adults only. There were 12497 adults in the sample. Of these, 10644 (85.2%) of them had travelled by bus before. Therefore, the analysis will be based on the 10644 sample size. The sample shows a gender distribution of 52.8% female and 47.2% male with and the average age of 38.13. 22.2% of the adults are 18 24 years, 46.9% are 25 44 years, 22.6% are 45 64 years and 8.4% are older t han 65 years. In terms of income, over 60% of the adults are from low income households, that is, the households annual income is less than 600ghc ($312 per 2012 cedi to dollar conversion rate). With respect to the marital status of the adults, 61.7% of the adults are married. The data includes some socioeconomic characteristics in addition to residential location of the adult and vehicle ownership. In capturing the perception of the adults with regards to trip purpose, the trip purposes were stratified a s people who m transport and included as explanatory variables as we hypothesize that they will have
81 The trip purposes/activities that were considered were work, he alth facility, food shop, other shop s (non food shop s ), post office, police stat ion and market. The descriptive statistics of adults based on the stratified trip purpose are shown in Table 4 1. Interaction variables were also created between the rural/urba n location variables and the vehicle ownership variables. Table 4 2 also presents descriptive statistics of the perception variables. Further, a combined variable of all the satisfaction levels were created to determine those who were satisfied with everyt hing and those who were satisfied with nothing. The aim of this is to understand the characteristics of the users who complain about everything and those who are utterly satisfied with everything. The full data descriptive of the variables considered for t his study are shown in Table 4 3 and Table 4 4. Modeling Methodology In Chapters 2 and 3, we described the travel behavior of the Ghanaian citizens with regards to their vehicle ownership and trip generation. However, due to the endogeneity issue with th e perceptions variables in the data, we excluded the perception variables from the vehicle ownership and trip generation models. These perception variables are further studied in this chapter to understand if public transportation can be used for all trip purposes and understand the effect of socio economic, location and other characteristics of the data on bus perceptions. In this chapter, we further discuss the modeling methodology used for analysis, its structure and assumption. Due to the dichotomous nature of the perception (dependent) variables, the binary logit models are employed for the analysis. The binary logit model predicts the probability that the observations fall into one of the two dichotomous dependent variables and it is a function of th e explanatory variables which
82 may be continuous or categorical. The perception variables (y) take on two values; 0 and 1, were: This study analysis six binary logit s based on the public transportation perceptions of adults who had ever use public transport (bus). The six binary logit models are based on these perceptions: Bus Route Convenience Bus Route C ondition Transport Availability Satisfaction Bus Frequency Schedule Satisfaction Overall Satisfaction with everything Overall Satisfaction with nothing The general function of the logit model is given by: Where The logit model assumes a logistic distrib ution, and the conditional probability that an adult is satisfied with any of the six outcomes variables is represented by the function below: The binary logit model howev er assumes a linear relationship between the logit of the outcome and the explanatory variables but not a linear relationship between the
83 dependent and independent variables. Another assumption of the binary logit model specification is the absence of mul ticollinearity among the explanatory variables. The binary logit models were developed to understand how the socio economic characteristics of an adult and other variables influence their perceptions about public transport. The six binary logit models we re structured and developed using the SPSS statistical software. The models developed were subject to several iterations by considering all the explanatory variables shown in Table 4 3 however, the final model only represents the explanatory variables wit h a t least a 95% confidence level. Results The binary logit models of the different satisfaction levels for all adults are presented in Tables 4 5 and 4 6. Table 4 5 presents the results o n the perceptions of all adults who had ever used a bus whereas Tabl e 4 6 shows the findings based on the overall satisfaction with using a bus. All the significant explanatory variables help understand the perception of a road user with regards to the satisfaction levels. T he effect of age on the levels of satisfaction i s only for adults between 18 and 44 years. Such adults compared to adults older than 65 years were more likely to be satisfied with the transport availability and bus frequency schedules. On the contrary, adults within this same age category were more like ly to be satisfied with nothing. This could possibly mean that the satisfaction for bus route and bus condition may have a higher influence in the combined satisfaction levels for such adults. It is also useful to acknowledge that the influence of gender a cross all levels of satisfaction was in significant. With regards to income, generally adults in households with income between 200 ghc and 600 ghc are less likely to be satisfied with the convenience of the bus route but
84 more likely to be satisfied with the bus condition and availability of transport. However, adults with higher household income s than the former have a higher dissatisfaction for the convenience of the bus route. Interestingly, adults from higher income households (income greater than 1200 ghc) have a higher overall satisfaction with using the bus than adults from households with middle income between 600ghc and 1000ghc. Additionally, such middle income household adults are generally satisfied with nothing comparing it to th e overall satisfa ction. A plausible explanation to these findings could be that higher income adults (representing 6.8% of the sample) rarely used the bus and were quite biased in rating their satisfaction compared to the middle income adults. The findings also show the s ignificant effect of ethnicity and regional location on f bus transport usage. It is important to note that, in Ghana, ethnicity can be linked to regional location s to a larger extent because most regions represent certain specific e thnic groups. Adults from the Brong Ahafo have the highest dissatisfaction in the convenience of the bus route, transporta tion availability and bus frequency schedule compared to adults in the other regions. Meanwhile, in terms of overall satisfaction, the relative effect of adults in the Brong Ahafo Region being satisfied with everything compared to being satisfied with nothing is positive. This could mean that the satisfaction for bus condition may have a higher influence in the combined satisfaction leve ls for such adults. Also, adults who live in a rural area with at least a four wheeler are generally satisfied with nothing although they may have a higher satisfaction for transport availability and bus frequency than bus route convenience. Adults living in rural areas without two wheelers show a positive correlation in terms of satisfaction for all and satisfaction for nothing although the
85 coefficient for satisfaction for all is slightly higher than satisfaction for nothing. Meanwhile on the different sa tisfaction levels, such adults are generally dissatisfied with transport availability and bus frequency. This could mean that such adults who are generally satisfied with nothing are the ones mostly dissatisfied with the transport availability and bus freq uency. In capturing the effect of satisfaction based on the activity purpose, adults who use the bus as their main transport mode for two or more activities are more satisfied with the bus condition and transport availability than adul t s who use the bus f or just one activity or purpose.
86 Figure 4 1 Perception ranking scale 2 Table 4 1 Bus as a Primary Mode by Activity Purpose Frequency Percent (%) Bus is never the main transport mode for any purp ose (Zero Purpose) 6246 58.7 Bus is the main transport mode for one purpose (One Purpose) 1546 14.5 Bus is the main transport mode for two purposes (Two Purposes) 1276 12 Bus is the main transport mode for three or more purposes (Three or More Purposes) 1576 14.80 Total 10644 100 Table 4 2 Proportion of the Perception of Adults on Public Transportation (Bus) Based on their Activity Purpose Number of Purposes for which Bus is the Primary Mode Sample Bus Route Convenient Bus Condition Satisfaction Transport Availability Satisfaction Bus Frequency Schedule Satisfaction Yes No Yes No Yes No Yes No Zero Purpose 6246 57% 43% 81% 19% 54% 46% 63% 37% One Purpose 1546 55% 45% 86% 14% 50% 50% 65% 35% Two Purposes 1276 50% 50% 82 % 18% 36% 64% 59% 41% Three or more Purposes 1576 53% 47% 78% 22% 37% 63% 60% 40% 2 Chowdhury, Subeh, et al. "Public transport users' and policy makers' perceptions of integrated public transport systems." Transport policy 61 (2018): 75 83.
87 Table 4 3 Descriptive Statistics of All Adults who have Ever Used a Bus Variables Proportion All Adults Adult Bus Users Sample 12497 10644 Gender Male 46.8 47.2 Female 53.2 52.8 Age Categories 18 24 years 23.5 22.2 25 44 years 45.4 46.9 45 64 years 22.3 22.6 > 65 years 8.8 8.4 Household monthly Income HH with monthly income less than 200 ghc 38.4 36.9 HH with monthly income between 200.01 ghc and 600 ghc 33.4 34.2 HH with monthly income between 600.01 ghc and 1000 ghc 12.2 12.6 HH with monthly income between 1000.01 ghc and 1200 ghc 3.7 4 HH with monthly income greater than 1200 ghc 6.3 6.8 Unknow n Income 6 5.5 Employment Status Employed 77.2 78.5 Unemployed 22.8 21.5 Ethnicity Akan 46.3 48.4 Ga/Dangme 6.8 5.9 Ewe 13.4 13.2 Mole Dagbani 19.8 19.3 Other ethnicity 12.9 12.5 Religion Christian 72.2 73.5 Islam 17.7 17.7 Traditional 5.2 4.1 Other religion 5 4.6 Marital Status Never married 26.7 25.9 Married 60.8 61.7 Divorce/Separated 5.6 5.8 Widowed 6.9 6.6 Residential Location and Vehicle Ownership Household member lives in an urban a rea with at least one four wheeler 4.3 4.4
88 Table 4 3. Continued Variables Proportion All Adults Adult Bus Users Household member lives in an urban area without a four wheeler 36.9 38.9 Household member lives in a rural area with at least one fou r wheeler 1.2 1.2 Household member lives in a rural area without a four wheeler 54.4 53.4 Household member lives in an urban area with at least one two wheeler 10.7 11.2 Household member lives in an urban area without a two wheeler 29 30.5 Household me mber lives in a rural area with at least one two wheeler 20.1 19.9 Household member lives in a rural area without a two wheeler 35.5 34.7 Regional Location Western 9.6 9.6 Central 6.3 6.5 Greater Accra 13.2 13.1 Volta 8.5 7.9 Eastern 11.2 1 0.3 Ashanti 19.9 21.1 Brong Ahafo 10.4 11.1 Northern 11.3 11.9 Upper east 6.6 5.6 Upper west 3.1 2.7 Note: Four wheelers is used to describe personal cars, buses and trucks
89 Table 4 4 Descriptive Statistics of Adul ts who have Ever Used a Bus based on their Overall Satisfaction Variables Proportion Not Satisfied Satisfied (All) Sample 961 2896 Gender Male 4 8.5 46 Female 51.5 54 Age Categories 18 24 years 22.30 23.40 25 44 years 50.60 45. 20 45 64 years 20.20 22.30 > 65 years 7.00 9.00 Household monthly Income HH with monthly income less than 200 ghc 44.50 34.00 HH with monthly income between 200.01 ghc and 600 ghc 35.50 32.10 HH with monthly income between 600.01 ghc and 10 00 ghc 11.20 14.60 HH with monthly income between 1000.01 ghc and 1200 ghc 2.00 4.90 HH with monthly income greater than 1200 ghc 3.30 10.40 Unknown Income 3.40 3.90 Employment Status Employed 78.60 77.10 Unemployed 21.40 22.90 Ethnicit y Akan 20.00 60.50 Ga/Dangme 4.40 4.00 Ewe 10.60 12.20 Mole Dagbani 35.70 14.20 Other ethnicity 28.50 8.00 Unknown 0.80 1.00 Religion Christian 52.90 76.10 Islam 33.10 18.10 Traditional 9.20 1.60
90 Table 4 4.Continued Variable s Proportion Not Satisfied Satisfied (All) Other religion 4.90 4.20 Marital Status Never married 22.00 28.80 Married 70.20 57.70 Divorce/Separated 2.80 6.50 Widowed 5.00 6.90 Residential Location and Vehicle Ownership Household m ember lives in an urban area with at least one four wheeler 1.50 6.10 Household member lives in an urban area without a four wheeler 23.90 52.00 Household member lives in a rural area with at least one four wheeler 0.20 0.90 Household member lives in a rural area without a four wheeler 71.10 40.80 Household member lives in an urban area with at least one two wheeler 10.60 15.20 Household member lives in an urban area without a two wheeler 14.60 41.90 Household member lives in a rural area with at leas t one two wheeler 41.50 14.50 Household member lives in a rural area without a two wheeler 29.80 27.20 Regional Location Western 2.30 8.70 Central 1.70 9.00 Greater Accra 11.80 9.50 Volta 12.10 5.90 Eastern 6.60 9.70 Ashanti 10.80 23.70 B rong Ahafo 3.60 18.80 Northern 31.00 11.80 Upper east 16.90 2.20 Upper west 3.30 0.70
91 Table 4 5 Binary Logit Models on Bus Perception Variables Bus Route Convenient Bus Condition Satisfaction Transport Availability Satisfa ction Bus Frequency Schedule Satisfaction Coeff. Sig. Coeff. Sig. Coeff. Sig. Coeff. Sig. Constant 2.519 0.000 1.292 0.000 2.286 0.000 1.077 0.000 Age Categories 18 24 years 0.164 0.012 0.141 0.032 25 44 years 0.124 0.014 0.124 0.024 Household monthly Income Income less than 200 ghc 0.307 0.001 0.242 0.001 Income between 200.01 ghc and 600 ghc 0.214 0.024 0.156 0.007 0.415 0.000 Income between 600.01 ghc and 100 0 ghc 0.237 0.025 0.170 0.041 0.359 0.000 Income between 1000.01 ghc and 1200 ghc 0.610 0.000 0.328 0.009 Income greater than 1200 ghc 0.407 0.001 Ethnicity Akan 0.529 0.000 0.738 0.000 0.693 0.0 00 0.177 0.004 Ga/Dangme 0.857 0.000 0.484 0.000 0.472 0.000 Ewe 1.025 0.000 0.589 0.000 0.426 0.000 0.478 0.000 Mole Dagbani 0.285 0.000 0.418 0.000 0.214 0.009 Marital Status Married 0.182 0.000 Residential Location and Vehicle Ownership Urban area with at least one four wheeler 0.270 0.012 Urban area without a four wheeler 0.253 0.021 Rural area with at least one four wheeler 0.868 0.000 2.256 0.035 Rural area without a four wheeler 1.151 0.000 2.437 0.020 0.602 0.003 Urban area with at least one two wheeler 1.411 0.000 0.947 0.000 0.271 0.002 Urban area without a two wheeler 1.329 0.000 0.139 0.035 0.721 0.000 Rural area with at least one two wheeler 0.419 0.000 2.171 0.038 0.548 0.009 Rural area without a two wheeler 2.175 0.038 0.667 0.001 Regional Location Western 0.677 0.000 0.591 0.000 1.424 0.000 0.352 0 .000
92 Table 4 5. Continued Variables Bus Route Convenient Bus Condition Satisfaction Transport Availability Satisfaction Bus Frequency Schedule Satisfaction Coeff. Sig. Coeff. Sig. Coeff. Sig. Coeff. Sig. Central 0.839 0.000 2.010 0.000 0.603 0.000 Greater A ccra 0.898 0.000 1.689 0.000 0.549 0.000 Volta 0.503 0.000 1.401 0.000 0.586 0.000 Eastern 0.480 0.000 1.565 0.000 Ashanti 0.949 0.000 1.265 0.000 0.304 0.000 Brong A hafo 1.690 0.000 2.185 0.000 1.532 0.000 Northern 0.863 0.000 1.171 0.000 1.538 0.000 Upper east 0.983 0.000 Activity P urpose Bus is never the main transport mode for any purpose 0.187 0.000 0.348 0.000 0.762 0.000 Bus is the main tran sport mode for one purpose 0.138 0.038 0.433 0.000 0.442 0.000 Bus is the main transport mode for two purposes/activities 0.281 0.005 0.188 0.026 0.184 0.011 Bus Frequency Schedule 0 15 mins 3.329 0.000 16 30 mins 1.941 0.000 31 45 mins 0.940 0.000 46 60 mins 0.721 0.000 >60 mins Ref Ref Ref Ref Ref Ref Ref Ref Note: Ref. = Reference
93 Table 4 6 Binary Logit Models on Overall Satisfaction with using Bus Variables Not Satisfied Satisfied (All) Coeff. Sig. Coeff. Sig. Constant 2.562 0.000 4.432 0.000 Age Categories 18 24 years 0.224 0.043 25 44 years 0.178 0.038 Household monthly I ncome Income less than 200 ghc 0.197 0.000 Income between 200.01 ghc and 600 ghc 0.322 0.000 Income between 600.01 ghc and 1000 ghc 0.388 0.001 0.177 0.014 Income greater than 1200 ghc 0.382 0.000 Ethnicity Akan 0. 937 0.000 0.865 0.000 Ga/Dangme 0.754 0.000 0.540 0.000 Ewe 1.167 0.000 0.836 0.000 Mole Dagbani 0.403 0.000 0.260 0.006 Marital Status Married 0.214 0.015 Residential Location and Vehicle Ownership Rural area wit h at least one four wheeler 2.014 0.005 Urban area with at least one two wheeler 1.549 0.000 Urban area without a two wheeler 1.540 0.000 Rural area with at least one two wheeler 0.822 0.000 Rural area without a two wheeler 0.635 0.00 0 0.705 0.000
94 Table 4 6. Continued Variables Not Satisfied Satisfied (All) Coeff. Sig. Coeff. Regional Location Western 0.935 0.000 1.132 0.000 Central 0.769 0.004 1.676 0.000 Greater accra 0.594 0.000 0.840 0.001 Volta 1.049 0.000 1. 075 0.000 Eastern 1.097 0.000 Ashanti 1.411 0.000 Brong ahafo 0.901 0.000 2.206 0.000 Northern 1.061 0.000 1.799 0.000 Upper east 1.536 0.000 0.769 0.005 Activity purpose Bus is never the main transport mode for any purpose 0 .429 0.000 0.432 0.000 Bus is the main transport mode for one purpose 0.600 0.000 0.261 0.004 Bus is the main transport mode for two purposes/activities 0.328 0.018 0.216 0.031
95 CHAPTER 5 SUMMARY AND CONCLUSION Summary Understanding the travel beh avior of road users is key in travel demand forecasting, however in most developing countries like Ghana there has been little contribution to understanding the travel pattern of road users. This study presents three different studies on understanding tr aveling behavior by using the 2012 Ghana Transport Indicator Database Survey. Apart from the Ghana Statistical Services which developed a report based on this survey, the disaggregate data has not been used for any research relating to transportation or tr ip generation as such. Therefore, results from this data serve as a useful contribution to the Ghana Statistical Services, Ministry of Transport and other governmental transportation agencies in developing national policies. The study first examines the t rip generation of all adults, considering both working and non working adults. Another key determinant in the study of travel demand is the understanding of household vehicle ownership. This is also important in making future traffic predictions. Ghana has been experiencing an increasing grown in vehicle ownership therefore there was the need to understand the factors that influence vehicle ownership. The second part of this study presents a vehicle ownership model for all households in Ghana. With the i ncrease in vehicle ownership and the consistent increase in traffic congestion in Ghana, it is needful to adopt a multimodal transport system to build a more sustainable transport system.
96 Public transportation is a key factor in helping to improve the tra nsportation system in any country. The third part of the study subtly tries to understand the public transportation perceptions of road users in Ghana. The main form of public transport used in the country is the bus therefore the perceptions are based o n bus usage. Common findings from the first part of the study shows that, gender, age, education level and residential location all have a significant impact on the number of trips generated by working and non working adults for the different trip modes. Household income also significantly impacts trips generation. Lower income workers make fewer car trips than high income workers the lower the income, the less likely the worker would make a car trip. The marital status of an adult also significantly af fects the car, bicycle, motorcycle and taxi trips a worker make s but only impacts the car, motorcycle and bicycle trips for non workers. From the vehicle ownership model, in addition to the effect of income, education and residential location on household vehicle ownership, the dynamic household structure significantly impacted the vehicle type a household was likely to own. Households with predominantly female adults are less likely to own at least a two wheeler compared to households with an equal propor tion of male to female adults. The general finding from the public transport users perception show s that the more adults use the bus for different activities the more they adjust to the bus conditions, availability, schedule and route and as such become more comfortable and satisfied with them. Also, all high income adults are generally satisfied with the bus transport system but are more likely to complain about the bus route.
97 The current forecasting model used in practice in Ghana is an aggregate mode l with few explanatory variables. In this study disaggregate (household and person level) models have been developed for both long term (vehicle ownership) and short term (trip generation) choices. These new models provide a significantly improved approa ch for examining the effects of socio economics changes on travel patterns. Research Implications Based on the challenges that were encountered in using the 2012 Transport Indicator Survey data, some suggestions and recommendations have been provided for performing future transport surveys. The first phase of the first ever nationwide household based transport survey was conducted in 2007 and the second phase in 2012 showing a 5 year data collection cycle. Equal cycles for collecting tran s port data on the same households should be encouraged in order to be able to conduct longitudinal surveys. Although the survey data provides metadata documenting the study procedure and manual, it is recommended to generate solid technical metadata providing thorough detai ls of each variable. The survey should collect data on daily trips by purpose made by persons/households to be able to generate origin destination trips whiles developing a trip generation model. More land use questions such as housing unit types, commerci al land use and agriculture should be included in the surveys to be able to encourage research on the influence of land use on transport demand models. Additionally, future survey s should include more travel characteristics like vehicle age and type, vehi cle miles travelled (VMT) and accurate travel time. With regards to perceptions about public transport, more opinion related questions on safety and transport characteristics should be included. Public transport (PT) was broadly referred to as busses, but, future surveys should include other forms of PT like passenger trains
98 and light rails. Data should also be collected on shared ride usage especially with the emergence of shared rides like uber which are very prevalent in developing countries like Ghana f PT show s that the more adults use the bus for different activities, the more they are satisfied with the bus. Also, the results showed that adults are more likely to complain about the bus route. Ther efore, improving bus routes will encourage more bus ridership even amongst adults with high income households. Increasing PT ridership also helps in controlling the pressure on the roadway infrastructure by private cars and may help curb congestion as well From the bus perception results, adults in the Brong Ahafo and Central regions of Ghana were mostly dissatisfied with transport availability and bus frequency, therefore much consideration should be placed on improving the PT in these regions especially their frequency of service and transport availability. All these measures are to make PT ridership palatable with the hope of increasing ridership. The trip generation model developed served as a basis for conducti ng future research on travel demand forec asting in Ghana, especially since there is no solid disaggregate trip generation model developed for the country. Some implication s for travel demand forecasting include s improving PT through regular evaluations and providing social benefit s or incentives for PT ridership. Also, a multimodal transport system should be encouraged, and future research should promote integrating land use and travel demand forecasting models.
99 These suggestions have been provided to help us make better research in travel behavio r studies and ultimately help us make future predictions for traffic congestion and develop more accurate transportation plans.
100 LIST OF REFERENCES Abane, Albert M. "Travel behaviour in Ghana: empirical observations from four metropolitan a reas." Journal of Transport Geography 19.2 (2011): 313 322. Ackaah, Williams, and David O. Adonteng. "Analysis of fatal road traffic crashes in Ghana." International journal of injury control and safety promotion 18.1 (2011): 21 27. Adogla Target Private Car Owners to Reduce Traffic Congestion private car owners to reduce traffic congestion consultanttarget private car owners to reduce traffic congestion consul tant/. Afolabi, Oluwaseyi Joseph, Adegbayi M. Hassan, and Luqman O. Age. "Behavioral pattern of commercial public transport passengers in Lagos metropolis." Journal of Sustainable Development of Transport and Logistics 2.1 (2017): 40 50. Agyemang, Ernest. "Mode choice for long distance trips: Evidence from the Greater Accra Metropolitan Area of Ghana." Journal of Transport Geography 64 (2017): 150 157. Amoh Gyimah, Richard, and Eric Nimako Aidoo. "Mode of transport to work by government employees in the Kum asi metropolis, Ghana." Journal of Transport Geography 31 (2013): 35 43. Arasan, V. Thamizh, V. R. Rengaraju, and KV Krishna Rao. "Trip characteristics of travelers without vehicles." Journal of transportation engineering 122.1 (1996): 76 81. Ashalatha, R. V. S. Manju, and Arun Baby Zacharia. "Mode choice behavior of commuters in Thiruvananthapuram City." Journal of Transportation Engineering 139.5 (2012): 494 502. Auditor PERFORMANCE AUDIT RE PORT OF THE AUDITOR -GENERAL ON ROAD SAFETY IN GHANA, 30 Sept. 2010. www.ghanaweb.com/GhanaHomePage/NewsArchive/Managing Traffic Congestion in Accra 195642. Baskin, Ariadn 2018, wedocs.unep.org/bitstream/handle/20.500.11822/25233/AfricaUsedVehicleRepor t.pdf?sequence=1&isAllowed=y.
101 Belwal, Rakesh, and Shweta Belwal. "Public transportation services in Oman: A study of public perceptions." Journal of Public Transportation 13.4 (2010): 1. Ben Akiva, Moshe E., Steven R. Lerman, and Steven R. Lerman. Discrete choice analysis: theory and application to travel demand. Vol. 9. MIT press, 1985. Bhat, Chandra R., and J essica Y. Guo. "A comprehensive analysis of built environment characteristics on household residential choice and auto ownership levels." Transportation Research Part B: Methodological 41.5 (2007): 506 526. Button, Kenneth, Ndoh Ngoe, and John Hine. "Model ling vehicle ownership and use in low income countries." Journal of Transport Economics and Policy (1993): 51 67. Calvo, Christina Malmberg. Case study on the role of women in rural transport: access of women to domestic facilities. Washington, DC: World B ank, 1994. behavioral adaptation to unreliability in public transportation." Transportation Research Record 2351.1 (2013): 153 162. Chowdhury, Subeh, and Avishai Ceder. "Defi nition of planned and unplanned transfer of public transport service and user decisions to use routes with transfers." Journal of public transportation 16.2 (2013): 1. Chowdhury, Subeh, et al. "Public transport users' and policy makers' perceptions of inte grated public transport systems." Transport policy 61 (2018): 75 83. Coleman, Albert. "Road Traffic Accidents in Ghana: A Public Health Concern, and a Call for Action in Ghana, (and the Sub Region)." Open Journal of Preventive Medicine 4.11 (2014): 822. Co oke, Edgar, Sarah Hague, and Andy McKay. "The Ghana poverty and inequality report: Using the 6th Ghana living standards survey." New York: UNICEF (2016). Currie, Graham, Alexa Delbosc, and Sarah Mahmoud. "Factors influencing young ersonal safety on public transport." Journal of public transportation16.1 (2013): 1. transit quality." Transport Policy 17.6 (2010): 388 397. Etminani Ghasrodashti, Roya, and Mahyar Ardeshiri. "Modeling travel behavior by the structural relationships between lifestyle, built environment and non working trips." Transportation Research Part A: Policy and Practice 78 (2015): 506 518. Modern Ghana, 14 Feb. 2017, www.modernghana.com/news/755613/the economic cost of traffic in accra.html.
102 Hart, Jennifer. Ghana on the go: African mobility in the age of motor transportation. Indiana University Press, 20 16. Haybatollahi, Mohammad, et al. "Neighbourhood preferences, active travel behaviour, and built environment: an exploratory study." Transportation research part F: traffic psychology and behaviour 29 (2015): 57 69. Hazen, Alyson, and John E. Ehiri. "Road traffic injuries: hidden epidemic in less developed countries." Journal of the National Medical Association 98.1 (2006): 73. Hesse, Christian A., and John B. Ofosu. "Epidemiology of road traffic accidents in Ghana." European Scientific Journal, ESJ 10.9 ( 2014). Ghanaweb.com, 18 Oct. 2011, www.ghanaweb.com/GhanaHomePage/NewsArchive/Traffic Congestion Problems in Accra Possible Solutions 221855. Hu, Xiaowei, Jian Wang, and Lei Wang. "Und erstanding the travel behavior of elderly people in the developing country: a case study of Changchun, China." Procedia Social and Behavioral Sciences 96 (2013): 873 880 Kumar, M., and K. V. Krishna Rao. "A stated preference study for a car ownership model in the context of developing countries." Transportation Planning and Technology 29.5 (2006): 409 425. Kumar, Pawan, S. Y. Kulkarni, and M. Parida. "Security perceptions of Delhi commuters at Metro bus interchange in multi modal perspective." Journal of tr ansportation security 4.4 (2011): 295 307. Lim, Kwang Kyun, and Sivaramakrishnan Srinivasan. "Comparative analysis of alternate econometric structures for trip generation models." Transportation Research Record: Journal of the Transportation Research Board 2254 (2011): 68 78. Loukaitou Sideris, Anastasia. environments: Case studies and best practices. No. FHWA CA MTI 09 2611. 2009. Matas, Anna, Jos Luis Raymond, and Jos Luis Roig. "Car ownership and access to job s in Spain." Transportation Research Part A: Policy and Practice 43.6 (2009): 607 617. Mohammed, Awal, Abdul Rahaman Haadi, and Ahmed Jamal Deen. "Identification of Risk Factors Involved in Road Accidents in Ghana: A Case Study of the Techiman Municipality ." International Journal of Statistics and Applications 5.1 (2015): 5 14.
103 MyJoyOnline.com, 19 May 2017, www.myjoyonline.com/news/2017/may 19th/bad roads traffic congestion cos t ghana billions yearly.php. Morse, Lawrence B., and Julian M. Benjamin. "Analysis of feeling of security on public transit among residents of small urban area." Transportation research record 1557.1 (1996): 28 31. National Road Safety Commission AFFIC CRASHES IN GHANA Statistics ( 2017 ): 29. Nyamuame, Godwin Y., et al. "Analysis of Road Traffic Accidents Trend in Ghana: Causing Factors and Preventive Measures." International Journal OF Engineering Sciences & Management Research 09 (2015). Or ( OICA ) http://www.oica.net/category/vehicles in use/ Perone, Jennifer S., and Lisa Tucker. An exploration of triangulation of methodologies: Quantitative and qualitative methodol ogy fusion in an investigation of perceptions of transit safety. No. NCTR 416 08.1. 2003. Peters, Deike. "Gender and transport in less developed countries: A background paper in preparation for CSD 9." London, Paper Commissioned by UNED Forum. 2001. Potogl ou, Dimitris, and Yusak Susilo. "Comparison of vehicle ownership models." Transportation Research Record: Journal of the Transportation Research Board 2076 (2008): 97 105. of slum residents in Nairobi, Kenya." Transport Reviews 30.5 (2010): 641 657. Schimek, Paul. "Household motor vehicle ownership and use: How much does residential density matter?" Transportation Research Record: Journal of the Transportation Research Boar d 1552 (1996): 120 125. Smith, M. N. "The number of cars worldwide is set to double by 2040." World Economic Forum, Geneva. 2016. Soltani, Ali, and Yousef Esmaeili Ivaki. "The influence of urban physical form on trip generation, evidence from metropolitan Shiraz, Iran." Indian Journal of Science and Technology 4.9 (2011): 1168 1174. Songsore, Jacob. "The urban transition in Ghana: Urbanization, national development and poverty reduction." University of Ghana, Legon Accra (2009). Takyi, Isaac K. "Trip genera tion analysis in a developing country context." Transportation Research Record 1285 (1990): 9 21.
104 Train, Kenneth, and Michael Lohrer. Vehicle ownership and usage: an integrated system of disaggregate demand models. Cambridge Systematics, Incorporated, 1982 UNECE. Used Vehicle Global Overview UNECE. 2017, www.unece.org/fileadmin/DAM/trans/doc/2017/itc/UNEP ITC_Background_Paper Used_Vehicle_Global_Overview.pdf. Van Exel, Nicolaas Jacob Arnold, and Piet Rietveld. "Perceptions of public transport travel time and their effect on choice sets among car drivers." Journal of Transport and Land Use 2.3/4 (2010): 75 86. Venter, Christoffel, Vera Vokolkova, and Jaroslav Michalek. "Gender, residential location, and household travel: Empirical findings from low income urban settlements in Durban, South Africa." Transport Reviews 27.6 (2007): 653 677. Walker, Joan, et al. "Travel demand models in the developing world: correcting for measurement errors." Transportation Letters 2.4 (2010): 231 243. World Health Organizatio n. Global status report on road safety 2015 World Health Organization, 2015. Ye, Runing, and Helena Titheridge. "Satisfaction with the commute: The role of travel mode choice, built environment and attitudes." Transportation Research Part D: Transport a nd Environment 52 (2017): 535 547.
105 BIOGRAPHICAL SKETCH Marian Ankomah was born in Accra, Ghana. She began her undergraduate studies at Kwame Nkrumah University of Science and Technology (KNUST) in Kumasi, Ghana. After the first year in KNUST, she tran sferred to Youngstown State University c ivil/ e nvironmental e ngineering in S c ivil e ngineering with a minor in Urban and Regional Planning at Univers ity of Florida. After earning her 2015, she continued with her doctoral degree as a research assistant. Her primary research interest is in traffic impact studies, traffic safety analysis, traffic operation analysis and travel dem and forecasting in developing countries, and qualitative and quantitative research methods. Her dissertation work focuses on the transportation demand patterns in Ghana by examining how the vehicle fleet composition of a household together with the househo ld and individual characteristics influence the travel behavior of road users. Marian grad uated and received her Ph.D. in civil e ngineering from the University of Florida in the summer of 2019.