|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
1 MODE CHOICE MODELS FOR AIRPORT ACCESS AND EGRESS TRIPS By BENJAMIN REIBACH A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DE GREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2013
2 201 3 Benjamin Reibach
3 To Lois and Paul
4 ACKNOWLEDGMENTS I would like to thank Dr. Siva Srinivasan for his mentorship and help while I was a member of the Transportation Graduate program at the University of Florida. Anytime I had a question, even with a simple answer, he was available to offer his knowledge and expertise. Throughout both of his classes he promoted a positive atmosphere which made it easy to learn and ask questions if I did not u nderstand the material properly. I would also like to thank Dr. Yafeng Yin and Dr. Ruth Steiner for serving as emphasize the importance of learning the material and strived to he lp students in any way possible. Dr. Steiner was also available to answer questions and lend a hand when asked about transportation planning in general. I would also like to thank my family and friends for their support throughout the graduate school process. Being far away from home and unable to see friends on a regular basis definitely has taken its toll on me. Maintaining constant communicatio n with both family and friends has given me the encouragement needed to complete my graduate program.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ ...... 4 LIST OF TABLES ................................ ................................ ................................ ................ 7 LIST OF FIGURES ................................ ................................ ................................ .............. 8 ABSTRACT ................................ ................................ ................................ .......................... 9 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ ........ 10 2 LITERATURE REVIEW ................................ ................................ .............................. 13 2.1 Summary of airport access mode choice: study regions and market segments ................................ ................................ ................................ ................. 13 2.2 Mode choice alternatives and modeling methodology ................................ ........ 16 2.2.1 Mode Alternatives ................................ ................................ ....................... 16 2.2.2 Mo del Methodology ................................ ................................ .................... 17 2.3 Empirical Findings ................................ ................................ ................................ 21 2.3.1 Impacts of the Characteristics of the Mode ................................ ............... 21 2.3.2 Impacts of the Characteristics of the Travel Party ................................ ..... 24 2.4 Summary ................................ ................................ ................................ ............... 26 3 DATA ................................ ................................ ................................ ........................... 27 3.1 Data Sources ................................ ................................ ................................ ........ 27 3.2 Data Processing ................................ ................................ ................................ ... 28 4 EMPIRICAL RESULTS ................................ ................................ ............................... 37 4.1 Resident Business ................................ ................................ ................................ 37 4.1.1 Resident Business Access ................................ ................................ ......... 38 4.1.2 Resident Business Egress ................................ ................................ .......... 39 4. 1.3 Resident Business Model Comparison ................................ ...................... 39 4.2 Resident Non Business ................................ ................................ ........................ 43 4.2.1 Resident Non Business Access ................................ ................................ 43 4.2.2 Resident Non Business Egress ................................ ................................ .. 44 4.2.3 Resident Non Business Model Comparison ................................ .............. 45 4.3 Non Resident Business ................................ ................................ ........................ 48 4.3.1 Non Resident Business Access ................................ ................................ 48 4.3.2 Non Resident Business Egress ................................ ................................ .. 49 4.3.3 Non Resident Business Model Comparison ................................ .............. 50 4.4 Non Resident Non Business ................................ ................................ ................ 5 3
6 4.4.1 Non Resident Non Business Access ................................ ......................... 53 4.4.2 Non Resident Non Business Egress ................................ .......................... 55 4.4.3 Non Resident Non Business Model Comparison ................................ ...... 55 5 SUMMARY AND CONCLUSIONS ................................ ................................ ............. 59 5.1 Model Discussion ................................ ................................ ................................ 59 5.2 Conclusions ................................ ................................ ................................ .......... 61 5.3 Future Research ................................ ................................ ................................ ... 62 LIST OF REFE RENCES ................................ ................................ ................................ ... 64 BIOGRAPHICAL SKETCH ................................ ................................ ................................ 66
7 LIST OF TABLES Table page 2 1 Overview of major models ................................ ................................ .................... 15 2 2 Model modes and structures ................................ ................................ ................ 20 2 3 Explanatory variables specific to the mode ................................ .......................... 23 2 4 Explanatory variables specific to travel party ................................ ....................... 25 3 1 Access mode shares from original and final sample ................................ ........... 30 3 2 Egress mode shares from original and final sample ................................ ............ 30 3 3 Sample shares of explanatory variables ................................ .............................. 33 4 1 Resident business access and egress mode cross tabulation ............................ 37 4 2 Resident business access model ................................ ................................ ......... 41 4 3 Resident business egress model ................................ ................................ .......... 42 4 4 Resident non business access and egress mode cross tabulation ..................... 43 4 5 Resident non business access model ................................ ................................ .. 46 4 6 Resident non business egress model ................................ ................................ .. 47 4 7 Non resident business access and egress mode cross tabulation ..................... 48 4 8 Non resident business access model ................................ ................................ ... 51 4 9 Non resident business egress model ................................ ................................ ... 52 4 10 Non resident non business access and egress mode cross tabulation .............. 53 4 11 Non resident non business access model ................................ ............................ 57 4 12 Non resident non business egress model ................................ ............................ 58
8 LIST OF FIGURES Figure page 3 1 GIS points of Oakland departures ................................ ................................ ........ 35 3 2 GIS points of San Francisco departures ................................ .............................. 35 3 3 GIS points of San Jose departures ................................ ................................ ....... 36
9 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science MODE CHOICE MODELS FOR AIRPORT ACCESS AND EGRESS TRIPS By Benjamin Reibach May 2013 Chair : Sivaramakrishnan Srinivasan Major: Civil Engineering Airports in major metropolitan cities attract large volumes of passenger traffic across a variety of modes including personal automobiles, rental cars, taxis, public transportation, and airport shuttles. Yet, there is relatively limited empirical research on studies capture the influence of traveler characteristics only to a limited extent. Further, practically all studies have examined only the choice of mode for access to the airport and not the choice of mode for egress from the airport. In this context, this study develops mode choice models for both access and egress trips for four main market segments (resident business, resident non business, non resident business, and non resident non business). Data were used from California to estimate the models. Seve ral factors such as time of day, day of the week of the trip, travel party composition, number of checked bags per person household income, whether costs are cover ed by the employer, duration of the long distance trip, and the distance of the access trip were all found to impact the choice of mode
10 CHAPTER 1 INTRODUCTION For calendar year 2011, the top ten airports in terms of total enplanements in the Unit ed States featured 262,671,999 total enplanements (FAA, 2012). While not all of these people will be directly accessing these airports due to some coming from connecting flights, it is safe to say there is a huge sum of people accessing the airports on a d aily basis. Modes available for access and egress vary from airport to airport as well. For example, the Atlanta airport had the following mode shares in 2005: 25% were dropped off, 31% parked their own car, 16% used a rental car, 8% used a taxi, 2% used a limousine, 10% used public transit, 4% used a commercial shuttle or van, 4% used a hotel courtesy shuttle, and less than 1% responded with some other mode (Gosling, 2008). Access by private vehicle, whether getting dropped off or parking, in addition to e ither a taxi or limousine service are available at almost all airports. The variation is mostly in the form of the transit modes. Understanding how passengers get to and from the airport is a vital part of the planning process for both the airport and the metropolitan region Choosing where to focus the efforts in planning ground transportation to help better serve travelers accessing the airport will in turn promote more users to choose that airport especially in multi airport regions Without a mode choi ce model, analyzing policies involving changes or improvements to the airport and to ground access transportation can prove to be very difficult (Gupta et al., 2008). If most users prefer one/two modes, planners can allocate more resources to maintain the high quality of access provided by these modes. On the other hand, if airport planners are looking to promote an alternate mode,
11 such as a dedicated airport rail line, they can use insights from the mode choice model to make this new mode attractive. Mode choice models for airport access trips are also important for the overall transportation demand forecasting for the metropolitan region. Based on a survey of 23 MPOs (Gosling 2008), 15 (65%) reported using a special generator sub model for air passenger trips. In addition to this, the rest of the MPOs treated the airport trips the same as regional trips or excluded them altogether (Gosling, 2008). This indicates that traditional transportation demand forecasting techniques often do not account for airport trips, or simply categorize them as special generators. However, given the large volume these trips, such simplified treatments in the planning process in inadequate. It is useful to acknowledge that demand forecasting models do include detailed mode ch oice models for other daily trips such as commute, escorting children, and shopping. However, these models cannot be directly transferred for use in the case f airport access because of several reasons. First, the possible modes available for airport acces s are significantly different from those used on a daily basis. For instance, trips than they are for daily urban travel. In addition, modes such as rental vehicles and shared ride vans are also popularly used for airport access. Second, the factors influencing the choice of mode for airport access trips can be very different from that determining mode choice for other daily trips. For example, number of bags and the len gth of the long distance trip (which determines the duration of parking and the associated costs) could critically impact the choice of mode for airport access but these are not relevant in the case of daily trips. Finally, mode choice for daily trips in a
12 demand forecasting model are analyzed for only (and all) residents of the region. Airport access trips are made by both residents and tourists. Further, not all the residents travel to the airports (for example low income people would be less likely to make such trips). In the light of the discussions above, the objective of this study is to build mode choice models for airport access and egress trips. It is useful to note that practically all studies in the past have examined only the choice of mode fo r access to the airport and not the choice of mode for egress from the airport. These models will capture the effects of a variety of factors with emphasis on the trip and traveler characteristics. Data from an air passenger survey conducted in three air ports in the San Francisco Bay Area will be used in this analysis. The general flow of this research involves assembling data from an airport passenger travel survey, often administered only to departing passengers, analyzing them to remove inconsis tencies, and building the appropriate models with proper explanatory variables. Many studies have been done in the past on this topic, and these are detailed in Chapter 2. Chapter 3 presents a detailed look at the data used in this study and the sample for mation procedure. Chapter 4 presents the models developed in this study. Lastly, Chapter 5 summarizes the work completed and offers additional ways to improve the models in the future.
13 CHAPTER 2 LITERATURE REVIEW This chapter presents a synthesis of the current state of the art in mode choice modeling for airport access trips. An overview of the major models is presented in Section 2.1 identifying the airports studied and the traveler segments modeled. Section 2.2 discusses the mode choice alternatives in cluded in the different models and the statistical modeling methodology. Next, the major empirical findings from these models are presented and discussed in Section 2.3. The chapter ends with an overall summary and identifies the contribution of the propos ed work. 2.1 Summary of airport access mode choice: study regions and market segments Twenty two different models that ha ve already been developed were reviewed. Table 2 1 summarizes each of the models and the market segmentations used. Each of the four ma rket segments is included in a column, and any alterations to these segments are noted for each model. For example, (Business only) and (Non Business only) for the IKIA model indicate the model does not account for residency status, only for the business a nd non business trip purpose classification. There are four main market segmentations used in mode choice mo deling for airport access trips: resident business, non resident business, resident non business, and non resident non business. Eight of the models use these exact four market segmentations, in addition to the SJC model call ing it a different name, referring to it as personal rather tha n non business. The IKIA, KIX, NY and KOR model s vary slightly, in that they only feature a business segment and a non business segment. The NY model is partially segmented to account for differences between residents and non residents. The GER1 and GER2 m odels feature the same six market segments, using domestic,
14 European, and international destinations for business trips and domestic, short European, long European, and international segments for non business trips. The ATH model uses domestic and internat ional segments for both business and non business. The LON model varies slightly from this in that it features both domestic and international trips segments for the resident segments but all of the non resident trips are grouped together The GER3 model c lassifies trips by the direction they are going, such as outward and return. The BAY model only used data for resident segments, looking at both business and non business travelers. The TIA, KKIA, and BWI models did not expressly state which market segment s were included in the model. For other models such as GER1, GER2, LON, and GER3, the trip destination and length were used rather than the residency classification.
15 Table 2 1. Overview of major models Model Airport ( s ) Included Year Author Resident Busines s Non Resident Business Resident Non Business Non Resident Non Business IKIA Imam Khomeini International Airport 2012 Mamdoohi, et al (Business only) (Non Business only) KIX Kansai International Airport 2012 Keumi, et al (Business only) (Pleasure) KKIA King Khaled International Airport 2011 Alhussein TIA Taoyuan International Airport 2011 Jou, et al BWI Baltimore Washington International 2008 Cirillo, et al NY New York Area Airports 2008 Gupta, et al. (Business only) (Non Business only) GER1 German Airports 2006 Gelhausen Domestic, European, Intercontinental Domestic, European 1 4, European 5+, Intercontinental KOR Gimpo, Daegu 2006 Choo, et al (Business only) (Non Business only) GER2 German Airports 2006 Gelhausen et al Domestic, European, Intercontinental Domestic, European 1 4, European 5+, Intercontinental HK Hong Kong International 2005 Tam et al X X X X CHI 1 2004 Resource Systems Group X X X X ATL 1 Atlanta Hartsfield 2003 Atlanta Regional Commission X X X X ATH Athens International 2002 Psaraki et al (Domestic & International) (Domestic & International) LON 1 London Area: LHR, GTW, STN, LTN 2002 Halcrow Group Limited (Domestic & International) X (Domestic & International) X TOR 1 Toronto Pearson 2002 Halcrow Group Limited X X X X SJC 1 San Jose 2002 Dowling Associates X X (Personal) (Personal) OAK 1 Oakland 2001 CCS Planning and Engineering X X X X POR 1 Portland 1997 1998 Cambridge Systematics and Portland Metro X X X X GER3 German Airports 1996 Bondzio Outward Return Outward Return BOS 1 Boston Logan 1996 Central Transportation Planning Staff X X X X MIA 1 Miami 1995 Gannet Fleming X X X X BAY Bay Area: OAK, SFO, SJC 1985 Harvey X X 1 ACRP Synthesis used as primary documentation for model
16 2.2 Mode choice alternatives and modeling methodology Table 2 2 contains the mode alternatives for each of the models being reviewed in addition to the model structure of each of the models. 2.2.1 Mode Alternatives A ll of the models except two use the main three alternatives in some form: the person or party drives themselves and parks at the airport, the person or party is dropped off at the airport, or the person or party uses a rental car. The IKIA, KIX and KKIA mo dels only offer two alternatives each. Some of these modes are restricted to one market segment, such as the CHI, ATL, LON, OAK, POR, and BAY models limiting the use of drive and self parking to residents. The GER3 model also restricts the self park option to departing passengers. This is reasonable as most non residents would not drive themselves to the airport and park, as they have no available vehicle to use. In addition, some but not all models restricted the usage of rental cars to visitors only. Whil e it may not be common, some business and wealthy non business travelers will use rental cars to access the airport. The transit options available at each airport usually dictate what goes into this mode alternative. Some airports have readily available pu blic transport options while some have none. Modes can also be grouped together depending upon market shares For example, the NY, TOR, OAK, and POR models all group together the taxi mode and the limousine mode. While these are not exactly the same mode the characteristics of each mode are quite similar. Both modes offer a door to door service with the same travel time. The only glaring difference between the two is the travel cost. Any modes that cannot be explicitly placed into one of the categories o r grouped with another are listed
17 under other. Some models considered other options as well. This included the IKIA, BWI, HK, CHI, and OAK models. The IKIA only considered two options, private and public, and thus these are grouped under the other category The BWI model was specifically focused on analyzing the use of cyber cars and shuttle options. For the HK model, airport shuttle buses and limousine services provided by hotels were considered other modes. The CHI model included both the other public and other private mode grouping. Modes that fell under the other private branch were limousine, hotel courtesy shuttle, and shared ride airport van service. Modes that fell under the other public branch were PACE buses, METRA train service, regional bus and c harter bus. 2.2.2 Model Methodology Table 2 2 shown below summarizes the available modes and structures of each model. An X in the column indicates that mode was available to be chosen. If the X is between two columns, those modes were combined within the model. Modes restricted to a given market segment are listed in parentheses. The Two main types of model structures have been featured in the previous studies are the Nested Logit model and the Multinomial Logit model. The Nested Logit structure is an enha nced version of the Multinomial Logit model that accounts for greater similarities among a subset of modes. Seven of the models featured the Nested Logit e xclusively, where five of the models featured the Multinomial Logit structure exclusively. Three stud ies used Binomial Logit one used Logistic Regression, and one us ed Binomial Logit Diversion all these are essentially multinomial logit models with only two mode alternatives. Two model s utilized both of the structures, with one using Multinomial Logit for the visitor segmentation and Nested Logit for the resident segmentation and the other varied dependent upon market segment
18 Understanding the nesting structure is of interest in the case of Nested Models The BWI model featured a two level nesting str ucture, with the top level containing four options: car driver, car passenger, transit, and taxi. The bottom level also featured four options, all of which fell under the transit option; these options included bus, metro and bus, metro and taxi, and shuttl e. For the NY model, because it was a joint airport choice and mode choice model, each of the mode choice alternatives fell into a nest underneath the airport choice alternatives. The GER1 and GER2 models used the same nesting structure, although one was f or a nested logit model and the other a neural logit model. There were two possible nests, separated into private modes and public modes; the private nest contained car, kiss and ride, rental car, and taxi, while the public nest contained bus, urban railwa y, and train. The CHI model featured different nesting structures for the business and non business segments. For business travelers, three nests were created: private modes, airport express train, and public modes. The non business featured two nests, wit h just private modes and public modes; the airport express train fell under the public mode nest in this case. The ATL model also featured different nests for, but this time for residents and non residents. The resident nest was split into private auto mod es and non auto modes. The non auto mode nest contained the transit and taxi mode choices. The non resident nest was split into three and feature dropped off, rental car, and the non auto nest again containing transit and taxi options. The LON model featur ed six different nesting structures depending upon market segmentation. Many of these separated out the auto modes from the transit modes into different nests. The BOS model featured only a Nested Logit structure for the resident market segment. The nestin g structure separated into two nests, one for door to door
19 modes and one for automobile modes. The MIA model used a nesting structure with two possible nests, one for group trips mainly containing transit options, and one for non group trips mainly contain ing private vehicle options. Overall, all nested models have generally grouped the private vehicle modes into a nest and transit options into another nest. The mixed logit structure of the TIA model featured a revealed preference choice and a stated prefe rence choice with the addition of a future mode. The currently available modes consisted of self driving, picked up, taxi, public bus, and HSR. The future additional mode being analyzed was the TIA MRT.
20 Table 2 2. Model m odes and s tructures Model Drive Self Park Dropped Off Rental Car Transit Taxi Limousine Shared ride Van Hotel Shuttle Other Model Structure IKIA Public, Private Binary Logit KIX Air, Bus Binary Logit KKIA Private Vehicle X Binary Logit TIA X X public bus, TIA MRT, HSR+TIA MRT X Mixed Logit BWI X bus, metro plus bus, metro plus taxi, shuttle X Shuttle, Cybercar Nested Logit NY X X X rail, local bus, charter bus X X Nested Logit GER1 X X X Urban Transit, Train, Bus X Neural Logit KOR X bus, subway X Logistic Regression GER2 X X X Urban Transit, Train, Bus X Nested Logit HK Private Vehicle AEL, Franchised Buses X X CHI X(resident) X X(non resident) Airport Express Train, Airport Express Bus, CTA Train X Public, Private Nested Logit ATL X(resident) X X(non resident) X X Nested Logit ATH X X X Public Bus, Tourist Bus X X MNL LON X(resident) X X(non resident) London Underground, Heathrow Express, Other Premium Rail, Bus/Coach X X Nested Logit TOR X X Airport Bus and Public together X Binomial logit diversion SJC Private Vehicle X Scheduled Airport Bus, Public Transit X X X(visitor only) MNL OAK X(resident) X(non resident) X Public X X X X MNL POR X(resident) X X(non resident) Scheduled Bus+Shared Ride Van, Light Rail, Express Bus X X MNL GER3 X(departing) X X(returning) X X MNL and Nested Logit BOS Nested Logit(Resident) MNL(Visitor) MIA X X X Premium, Local X X X X Nested Logit BAY X(resident) X X(non resident) Public, Airporter X MNL
21 2.3 Empirical Findings While reviewing the previous models, it is important to pay attention to similarities among them. Looking at the model results and the sign (positive or negative) on each coefficient can reveal much about what the model means. A positive sign means that the corresponding factor increases the utility of the alternative making it more attractive 2.3.1 Impacts of the Characteristics of the Mode Table 2 3 shown below summarizes the explanatory variables of each model specific to the modes. Unless otherwise stated, each variable applies to every mode available to the decision maker. For example, for the IKIA model, the peak travel time and travel cost are included as a parameter for both available modes. The ATL model features transit in vehicle travel time and off peak highway time as separate parameters dependent upon the mode utilized. The wait ti me for out of vehicle travel time in the ATL model also only applies to the transit mode. Every model reviewed features some version of travel time and travel cost. How each model uses these variables is where they vary In general, there are two typ es of travel time: in vehicle travel time and out of vehicle travel time. Out of vehicle travel time can be at the beginning of the trip, such as walking from home to the train station, or at the end of the trip, such as walking from the parking lot to the terminal. Travel time home. They want to get where they are going the fastest, and this plays a big role. Travel cost is also featured in every model reviewed, and agai n used in different ways from model to model. Sixteen of the models assign a travel cost to every mode alternative available. In addition to this, the POR model adds a cost to the drop off mode for vehicle operation. The ATL model uses two cost components for private
22 vehicle, operating and parking, and also uses travel costs for transit fare and taxi fares. Four models also feature a version of a variable that accounts for transfers. This applies only to the transit options for each model. While none of the models explicitly contain a variable to account for time of day, the travel time and travel cost can often implicitly contain this information. Whether or not the trip is made during peak or off peak times can affect both the travel time and travel cost. When analyzing effects of each of these, a higher travel time and travel cost tends to make a mode less attractive.
23 Table 2 3. Explanatory v ariables s pecific to the m ode Model IVTT OVTT Travel Cost Number of transfers IKIA Peak Travel Time Both Modes KIX Travel time, IVTT Ticket Cost KKIA Total Travel Time TIA IVTT OVTT Self driving ,TIA MRT, HSR BWI All Modes All Modes NY All Modes, Rail Transit, Bus Transit (Rail and Bus weighted by 2) All Modes GER1 Total travel time All Modes KOR Total travel time All Modes GER2 Total travel time All Modes HK All Modes CHI Weighted travel time All Modes Use of Intermediate Station ATL Transit, Off Peak Highway Time Transit walk time, Transit wait time Private vehicle (operating, parking),Transit fare, Taxi fare ATH All Modes LON IVTT+Walk time Wait time/headway All Modes X TOR All Modes Wait time/headway All Modes SJC Private Vehicle, Rail Transit, Bus Transit Wait time/headway, Walk distance All Modes OAK Private Vehicle, Rail Transit, Bus Transit Wait time/headway, Walk distance All Modes POR Total travel time(IVTT, wait, On airport time) All Modes, Drop off vehicle operating cost GER3 Total travel time All Modes X BOS All Modes Auto access time to transit, Out of vehicle time All Modes X MIA All Modes Wait time + terminal time All Modes BAY Total travel time(all modes) All Modes
24 2.3.2 Impacts of the Characteristics of the Travel Party Table 2 4 shown below summarizes the explanatory variables specific to the travel party for each of the models. As can be seen in the table, there are relatively few of these considered in any of the models. Looking at the characteristics of the tra vel party variables shows a lot of variance from one model to the next. There is not an established set of standard variables of this type that should be included in each model. Some of them do have a higher prio rity for inclusion however. Ten of the model s feature the household income variable. Higher income groups tend to choose more expensive modes, where lower income groups tend to choose cheaper modes such as public transit offerings. The other variable featured in multiple models is the number of chec ked bags When travelers arrive at the airport, they do not want to carry luggage all over the place. The more bags a person is traveling with tend to sway them towards a private option because carrying a lot of bags on public transit options can be very d ifficult. The BOS model features a dummy variable for whether or not the origin of the trip is a place of residence or not. Three models include a variable related to the gender of the traveler. The IKIA model concludes that men are more likely to choose p ublic transit than women. This seems intuitively correct, especially if the woman is traveling alone.
25 Table 2 4. Explanatory variables s pecific to t ravel p arty Model Driving distance Household income Air party size Number of Checked Bags Employer pays cost Flights/year Gender IKIA X X KIX X KKIA X TIA Past modes BWI NY X X X GER1 KOR X GER2 HK Times at Airport X CHI ATL ATH LON X TOR X X SJC X OAK X POR X GER3 X X BOS X X X X X MIA BAY X
26 2.4 Summary This chapter has presented the current state of the art in mode choice modeling for airport access trips. Two major issues emerge. First, none of the studies have examined the factors affecting the choice of mode for airport egress trips. Second, the incorporation of trip and traveler characteristics in the models is limited. This study seeks to address thes e two issues. Models for both access and egress trips are built for the four major market segments. All these models will include an extensive list of trip and traveler characteristics.
27 CHAPTER 3 DATA This chapter describes the data sources and the samp le formation procedure. The characteristics of the final analysis sample are also presented. 3.1 Data Sources The primary data used in this study come from the MTC Air Passenger Survey from the years 2001 and 2002 and is available for download from the MTC website (MTC) The survey is administered on roughly five major airports. Due to the tragic events of September 11 th 2001, which fell directly in and continued again using the same method the following year. The surveys were administered in the months of August and September of both years and on multiple days of the week including both weekdays and weekends at each airport. Whole flights were chosen through a lengthy process to be randomly sampled, and each t raveler on the chosen flights were given the survey while waiting for their departure flight. Those travelers who arrived at one of the airports and are departing on a connecting flight are exclu ded from being given a survey. The respondent ha d multiple options to return the survey, including filling it out on the spot and returning to the sampler, or mailing it upon completion of their travel. The full survey of 23 questions consisted of some abo ut the flight being taken such as origin and destination and some about the traveler or travel party characteristics including home location, household income, and number of checked and carry on bags they are traveling with In addition to these there are multiple questions about the mode used to arrive at and depart from the airport including how or why the person chose the mode they did. The survey listed the
28 following modes as available options for selection: private vehicle, rental vehicle, shuttle bus from train, regular transit bus, scheduled bus to airport, taxicab, hotel/motel courtesy shuttle, pre arranged limousine, pre arranged shared ride van, chartered bus, and by some other mode with the option to list it. There are more detailed questions abo ut each mode as well, such as how the traveler accessed the transit option, how many days they parked at the airport, and whether or not the cost of the access mode The raw data consisted of 5930 surveys in 20 01 and 8861 surveys in 2002 For 2001, 29.2% of travelers came from Oakland, 43.5% from San Francisco, and 27.3 percent from San Jose. For 2002, 27.3% came from Oakland, 41.6% from San Francisco, and 31.2% from San Jose. A secondary set of data used in thi s study is relate d to the inter zonal travel characteristics for the San Francisco Bay Area. Travel times and costs (from network skims) are available by mode and time of day. This data consists of 1099 TAZs. 3.2 Data Processing The data from the Air Passe nger Surveys were subject to substantial processing to arrive at the final analysis sample. To promote ease of work and to increase the sample size, data from both years of the survey were combined into one dataset. The first cleaning step was to remove du plicate cases. Those respondents who answered the question of how many expected questionnaires in your party as one were always retained. If the respondent had answered with a value greater than five, they were removed from the sample. If the answer to thi s question was two through five, the cases were sorted through to remove duplicates. A duplicate case indicated that more than one person from the same travel party had responded to the survey, and the additional
29 cases were removed to ensure each travel pa rty was only considered once. If a duplicate case was found, the case with the lowest sample number was retained, and all others removed. Following this step, the sample included all cases which had answered the question with one through five and did not h ave an easily identifiable duplicate case as well. Following the removal of these duplicate cases, the samples needed to be sorted into the four respective market segments. Two questions from the survey were used to accomplish this split. The first questi on was the trip purpose, either business, non business, or missing. If the trip purpose was missing, the case was removed from the dataset. The second question from the survey used was whether or not this was the home end of the trip. The home end was cons idered any origin within roughly 500 miles of the origin airport, and the origin airport not being further away than the destination airport. The case could have three possible values: yes this is the home end, no this is not the home end, or missing. Case s missing values were again removed from the dataset. Combining these two questions of trip purpose and home end allows you to separate the data into resident business, resident non business, non resident business, and non resident non business market segm ents. The survey questionnaire had eleven access mode options to choose from, but not all of these had a significant sample share to be included. In the survey, the private vehicle option was grouped together. This had to be separated into two options: the person was driven to the airport and dropped off, or the person drove to the airport and parked. This was done by using the curb drop off and parking status questions also from the survey. In addition to this, the three transit options were grouped togeth er into
30 one transit mode. These three options were shuttle bus from train, regular transit bus, and scheduled bus to airport. To promote consistency, the mode options of limousine and taxicab were also grouped together, as they have been in the egress mode option as well. Residents who chose the hotel shuttle mode option were removed due to a small sample size. Only samples that contained both the access mode and the egress mode were retained. The hotel shuttle was also not available as an egress mode in th e survey, so it is not included for any market segment. The modes listed in Table 3 1 and Table 3 2 are the final sample shares and were the only available modes considered in the models. Table 3 1. Access mode shares from original and final sample Access Mode Original Sample RB RNB NRB NRNB Final Sample Drive + Drop 35.6 30.8 45.0 13.3 46.9 37.7 Drive + Park 21.3 44.8 28.1 X X 19.2 Rental Car 18.1 2.5 1.3 52.4 30.3 17.4 Transit 7.6 4.8 11.6 6.7 7.0 8.1 Taxi+Limo 9.2 13.8 8.5 17.2 6.0 10.2 Hotel Shuttle 2.7 X X 5.1 3.6 2.1 Shared Ride 4.7 3.3 5.5 5.3 6.2 5.2 Note: RB is Resident Business; RNB is Reside nt Non Business; NRB is Non Resident Business ; NRNB is Non Resident Non Business Table 3 2. Egress mode shares from original and final sample Egress Mode Original Sample RB RNB NRB NRNB Final Sample Picked Up 40.7 30.60 44.7 18.39 50.2 40.5 Parked Vehicle 18.2 42.93 28.1 X X 17.5 Rental Car 15.1 2.59 1.0 45.93 26.1 15.1 Transit 7.9 4.95 10.8 8.00 8.05 8.6 Taxi+Limo 12 16.51 10.5 20.33 8.69 12.8 Shared Ride 5.2 2.42 4.9 7.35 6.99 5.5 Note: RB is Resident Business; RNB is Reside nt Non Business; NRB is Non Resident Business ; NRNB is Non Resident Non Business Further cleaning was done to ensure each case included the explanatory variables to be used in the analysis. Some of these variables were re coded using questions
31 directly from the survey. One of these variables was the household income variable. The surve y responders had a choice of ten different income levels to choose from. Grouping these together created the final four income levels as follows: Low Income ($0 $60,000), Medium Income ($60,000 $125,000), High Income ($125,000+), and Missing Income. Other variables that needed to be re classified included the arrival and departure times, the number of adults and children in the travel party, the number of checked bags, the day of the week, and the destination. The arrival and departure times were separated into six time groups. The number of children was grouped into 0, 1, 2, and 3+. To notice effects of adding a single adult, the number of adults were categorized as 2 or more, 3 or more, and 4 or more. Gender was only used for parties with a single adult tr aveling. Gender would be categorized as male or female only if the number of adults in the party was 1. The number of checked bags was grouped into 0, 1, 2, 3, 4, and 5 or greater. The day of the week was either a weekend or a weekday which could be determ ined by the date of the travel in the survey and was only used for the access model. Due to inconsistencies involving the length of the trip, the day of the week of the return trip could not be estimated. The destinations were classified as either domestic or international from the destination airport given in the survey. The origin location was different for the different market segments. As they do not live in the area, non residents could not be coming from their own home. Non residents could be coming f rom a hotel origin, place of business origin, or other origin type. Residents could either be coming from their own home or other origin type. Origin locations were only used for the access trip due to the egress location being unknown. Sample shares for e ach of these variables are shown in the table below. Cases with missing values for
32 any of these variables were removed from the dataset. Table 3 3 details the sample shares for each market segment for the explanatory variables used in the analysis. Categor ical variables are listed by percentage, while continuous variables are listed as an average. Arrival times were generally distributed throughout the day, but there is definitely a greater amount in the morning than the evening. The opposite can be said of departure times, where the majority was in the evening rather than the morning. The majority of all trips are made on weekdays, which makes sense due to sampling days being more readily available. Non business travelers were more likely to be going to an international destination than business travelers. Non business travelers also tend to have longer trips than their business counterparts. For single adult parties, males are more likely to have responded when they are on a business trip, and females are m ore likely to have responded for non business trip purposes. The number of adults and the number of children within the travel parties were also larger for non business trips than for business trips. The number of checked bags is divided by the total party size to give checked bags per person. Business travelers tend to travel with less checked baggage than their non business counterparts, which could be due to the fact their trips are shorter. Business travelers also tend to carry smaller bags and not have to wait for their checked bags to arrive. Business travelers have a higher portion in the high income level, while non business travelers have a higher portion in the low income level. Costs could only be reimbursed if the traveler is on a business trip, so this variable does not apply to non business travelers. An X is placed in the table where the variable does not apply to a given market segment.
33 Table 3 3. Sample s hares of explanatory variables Variable Reisdent Business Resident Non Business Non Resident Business Non Resident Non Business Whole Sample Single Male 58.3 24.0 52.1 20.2 34.1 Single Female 25.6 30.7 22.5 31.7 28.7 2+ Adults 16.1 45.3 25.4 38.1 37.3 3+ Adults 3.2 9.5 5.9 5.5 7.9 4+ Adults 1.3 4.3 2.8 4.5 3.6 Children 0.16 0.29 0.15 0.29 0.24 Low Income 6.8 20.9 14.6 30.3 19.7 Medium Income 31.6 32.9 37.9 30.2 32.7 High Income 46.2 26.4 32.7 18.1 29.0 Missing Income 15.4 19.8 14.9 21.4 18.6 Costs Covered 67.5 X 66.3 X 67.0 Costs Not Covered 32.5 X 33.7 X 33.0 Trips in Last Year 10.6 4.1 5.3 2.4 5.1 International 7.6 14.6 5.6 8.8 10.3 Domestic 92.4 85.4 94.4 91.2 89.7 Bags Per Person 0.67 0.68 0.65 0.84 0.70 Days Away 6.95 17.3 X X 13.83 Days Here X X 4.54 7.93 6.73 Arr 0:00 8:00 41.9 30.5 19.1 19.5 27.9 Arr 8:00 20:00 53.4 62.9 75.3 73.4 65.8 Arr 20:00 24:00 4.7 6.6 5.6 7.1 6.2 Dep 00:00 8:00 3.8 5.9 5.9 4.9 5.2 Dep 8:00 20:00 63.1 66.1 76.1 74.9 69.5 Dep 20:00 24:00 33.1 27.9 18.0 20.3 25.3 Own Home Origin 77.2 81.0 X X 47.5 s House Origin X X X 53.1 19.6 Place of Business Origin X X 30.9 X 11.6 Hotel Origin X X 51.9 35.7 18.8 Other Origin 22.8 19.0 17.2 11.2 2.6 Inside TAZ 92.3 91.5 91.1 85.2 89.85 Inside TAZ DA Time 36.35 40.74 31.64 39.53 38.03 Inside TAZ TR Time 69.51 69.22 64.31 65.63 67.52 Outside TAZ Distance 51.33 57.10 71.34 62.92 60.48 Weekend 20.0 25.0 17.3 36.8 26.1 Weekday 80.0 75.0 82.7 63.2 73.9
34 The next step involve d the determination of the transportation system characteristics by mode for the trips to and from the airport. Using GIS and the x coordinate and y coordinate of each origin location along with known TAZ data for the Bay Area, each case can be assigned to an origin TAZ. The geo coded (latitude/longitude) location of the trip origins was known for each of the airport access trips from the survey. These were mapped to the Traffic Analysis Zones (TAZs) centroid locations The location of the airports (TAZs) is known as well. Using this information and an available Level of Service table, travel charac teristics for each origin and destination pair can be added to the dataset through a match process. Travel characteristics added during the match process included the peak drive alone travel time and the peak walk access transit travel time In addition to matching each case to their origin, the straight line distance from origin location to the airport location can be mapped. Roughly ten percent of all cases fell outside of the 1099 TAZs. Travel times were only used for cases inside of the TAZ range by usin g a dummy variable to account for whether the case was inside or outside of the range. For cases outside of the TAZ range, the straight line distance from the origin location to the departure airport was used. Both the travel time and distance were only us ed for the access model, as the destination of the egress trip was unknown. Figure 3 1 through 3 3 shows the GIS points of the origin locations for each of the three airports. Both points within the TAZ range and outside the TAZ range are shown. The shaded area represents the TAZ range and is consistent across each figure.
35 Figure 3 1. GIS points of Oakland departures Figure 3 2. GIS points of San Francisco departures
36 Figure 3 3. GIS points of San Jose departures
37 CHAPTER 4 EMPIRICAL RESULTS This chapter presents each of the models formulated in this study in addition to a discussion about each model. The study resulted in eight different models being created over four different market segments. Both access and egress models were formulated for the resident business, resident non business, non resident business, and non resident non business travelers market segments. 4.1 Resident Business The final sample for the resident business market segment included 1817 cases. The full models for both the acc ess and egress modes are shown in Table 4 2 and 4 3 respectively. Table 4 1 shown below is a cross tabulation of the access mode and egress mode chosen sample shares. Table 4 1. Resident business access and egress mode cross tabulation Egress Mode Picked up Parked Car Rental Transit Taxi/Limo Shared Ride Drive Drop 23.06 1.98 0.33 0.94 4.02 0.50 Drive Park 0.00 44.80 0.00 0.00 0.00 0.00 Access Mode Rental 0.17 0.17 1.87 0.00 0.28 0.00 Transit 0.83 0.11 0.06 3.30 0.50 0.06 Taxi/Limo 2.15 0.00 0.22 0.22 11.01 0.17 Shared Ride 1.21 0.06 0.06 0.28 0.11 1.60 The majority of travelers use the same mode for both access and egress, but other modes are used as well. Nearly fifteen percent chose a different access and egress mode. If people are dropped off and choose a different egress mode, they are most like to choose the taxi/limo option to get home. If people use a taxi/limo to access the airport and switch for the egress trip, they are most likely to be picked up for their egress trip.
38 4.1. 1 Resident Business Access This model contained six available modes to be chosen from. These six modes were drive and drop, drive and park, rental vehicle, transit, taxi/limo, and shared ride. Single male travelers are less likely to choose the drive and d rop option, while single female riders are more likely to choose the shared ride option. Parties of two or more are likely to choose the drive and park option, but parties of three or more make this a less likely option. People in the low income category a re more likely to choose the transit option, which is generally the cheapest of all the available access modes. Travelers in the high income category are more likely to choose the drive and park option, which dependent upon length of stay, is often the mos t expensive available mode. If the costs are going to be reimbursed, the traveler is more likely to choose the taxi limo option. This provides them with a door to door service, and cost is not going to affect their decision. People going on longer trips ar e less likely to drive and park. Each additional day adds to the total cost, so people are more likely to choose a different origin was within the TAZ range. Those inside ar e less likely to choose the drive and park and rental options and are more likely to choose the transit option. For travelers arriving from an origin within the TAZ range, the longer their driving time the more likely they are to choose the drive and park option. In addition, the longer their driving time the less likely they are to choose the taxi/limo option. The cost of this option is usually based on distance, and a longer driving time usually means a longer distance, so this mode becomes very expensive The same can be said for those arriving from an origin outside of the TAZ range, the longer the distance, the less likely they are to choose the taxi/limo option. Travelers arriving on a weekday are less likely to choose the drive and
3 9 drop option. For mu ch of the day, people available to drive them would most likely be working. 4.1.2 Resident Business Egress This model also contained six available modes to be chosen from. These six modes were drive and drop (picked up), drive and park (parked vehicle), re ntal car, transit, taxi/limo, and shared ride. Single male travelers are less likely to choose the drive and drop option, while single female travelers are more likely to choose the shared ride option. For single female travelers, the comfort of not travel ing to the airport alone can be reflected in this. Groups of four or more adult travelers are likely to choose the rental option. Large groups traveling on a business trip like to stay together, and the ease and convenience of renting a car can accommodate this. Lower income people tend to choose the transit option, most likely because it is the cheapest option of the six. Travelers who have their costs covered are more likely to choose the taxi/limo option. This option gives them the most flexibility being a door to door service, and they do not have to worry about the cost being high. Travelers returning from longer trips prefer to use the renal option and prefer not to use the drive and park option. The longer a trip is, the more expensive the drive and p ark option becomes, so they seek out a cheaper alternative. Those passengers departing between 8:00AM and 8:00PM are less likely to be picked up. Between these hours is when most people are normally working, so they would not be able to pick the returning passenger up. 4.1.3 Resident Business Model Comparison The access and egress models for the resident business market segment offer both similarities and differences. Certain variables were only applied to the access mode due to data limitations. Single mal e travelers are less likely to choose the drive
40 and drop mode for both access and egress, while single female travelers are more likely to choose the shared ride mode for both directions. Parties of two or more are likely to choose the drive and park optio n for both directions, while parties of three or more are less likely to choose this option. Low income travelers are more likely to choose the transit option in both directions as this is often the cheapest available mode. For the access mode, the more tr ips in the last year from the origin airport, the less likely they are to choose the shared ride mode. For the egress mode, they are less likely to choose the drive and drop mode with the addition of more flights. For the access mode, the arrival and depar ture time did not affect which mode the traveler chose. For the egress mode, a departure time between the hours of eight and eight made them less likely to choose the drive and drop option, as the main people who could pick them up during these times were most likely working. Variables that were not significant in either model included the trip being to an international destination and the number of checked bags per person.
41 Table 4 2. Resident business access model Variable Drive Drop Drive Park Rental Transit Taxi Limo Shared Ride Single Male .374 ( 3.095) Single Female .60 (2.087) 2+ Adults .426 (2.449) 3+ Adults 1.53 ( 3.89) 4+ Adults Children Low Income 1.10 (3.643) Medium Income High Income .243 (2.18) Missing Income Costs Covered 1.45 (11.29) 1.06 (3.156) 2.56 (10.483) Costs Not Covered Trips in Last Year .013 ( 2.526) .062 ( 2.79) International Domestic Bags Days Away .426 ( 13.573) .172 (2.199) OAK .582 (4.237) .588 (1.753) 1.23 (3.557) .998 (1.971) SFO 1.77 (5.695) 1.3 (7.817) 2.40 (5.734) SJC Arr 0:00 8:00 Arr 8:00 20:00 Arr 20:00 24:00 Dep 00:00 8:00 Dep 8:00 20:00 Dep 20:00 24:00 Own Home Origin .284 (2.031) Other Origin Inside TAZ 1.17 ( 4.781) .989 ( 2.213) 1.26 (1.691) Inside TAZ DA Time .015 (5.868) .024 ( 5.252) Inside TAZ TR Time Outside TAZ Distance .039 ( 3.199) Weekend Weekday .293 ( 2.034) 2001 .203 (1.772) .739 (3.198) .88 (3.153) 2002 Constant .532 (1.599) 3.45 ( 5.495) 4.94 ( 6.074) 2.74 ( 8.862) 4.10 ( 8.453) Sample Size 1817 LL Convergence 1988.512 LL Constants 2446.108 Adj Rsq 0.18384
42 Table 4 3. Resident business egress model Variable Drive Drop Drive Park Rental Transit Taxi Limo Shared Ride Single Male .561 ( 4.586) Single Female .661 (1.978) 2+ Adults .624 (3.663) 3+ Adults .769 ( 2.172) 4+ Adults 1.46 (1.854) Children Low Income .436 (1.926) 1.25 (3.856) .88 (1.821) Medium Income High Income Missing Income Costs Covered 1.39 (10.992) 1.12 (3.294) 1.54 (8.861) Costs Not Covered Trips in Last Year .028 ( 4.732) International Domestic Bags Per Person Days Away .389 ( 13.345) .209 (2.661) OAK .782 (5.576) 1.91 (5.272) .459 ( 2.019) 2.17 (2.745) SFO 1.63 (4.619) .454 (3.089) 3.12 (4.253) SJC Arr 0:00 8:00 Arr 8:00 20:00 Arr 20:00 24:00 Dep 0:00 8:00 Dep 8:00 20:00 .227 ( 1.922) Dep 20:00 24:00 2001 .468 (2.054) .546 (1.7) 2002 Constant .180 ( 1.001) 4.64 ( 9.662) 4.07 ( 11.55) 2.34 ( 11.581) 5.95 ( 7.911) Sample Size 1817 LL Convergence 2053.695 LL Constants 2409.8866 Adj Rsq 0.14508
43 4.2 Resident Non Business The final sample for the resident non business market segment included 3353 cases. The full models for both the access and egress modes are shown in table 4 5 and 4 6 respectively. Table 4 4 shown below is a cross tabulation of the access mode and egress mode chosen sample shares. Table 4 4. Resident non business access and egress mode cross tab ulation Egress Mode Picked Up Parked Car Rental Transit Taxi/Limo Shared Ride Drive Drop 35.76 3.28 0.21 1.67 3.28 0.81 Drive Park 0.00 28.06 0.00 0.00 0.00 0.00 Access Mode Rental 0.39 0.03 0.57 0.09 0.18 0.09 Transit 1.97 0.45 0.00 7.96 0.51 0.72 Taxi/Limo 2.03 0.00 0.15 0.27 5.73 0.30 Shared Ride 1.31 0.06 0.06 0.39 0.81 2.89 The majority of travelers again choose the same mode for both access and egress. Nearly twenty percent however choose a different access and egress mode. The largest percent of those switching were dropped off for the access, but chose transit or taxi limo for the egress mode. 4.2.1 Resident Non Business Access This model contained six modes to be chosen from. These six modes were drive and drop, drive and park rental car, transit, taxi/limo, and shared ride. Travelers with children prefer not to use the transit option. Children can get restless and agitated on long trips, and often transit options have long travel times. Parents would prefer to have more direc t access to the airport than deal with children on transit options. Travelers from the low income group tend to prefer the transit options, while travelers from the high income group tend to prefer the taxi/limo option. These options tend to be among
44 the c heapest and most expensive respectively, so intuitively this seems correct. The more checked bags the person is carrying, the less likely they are to choose the transit option. Carrying many bags on a train or bus is often very difficult, especially if you have to transfer. Instead, people with many checked bags prefer the taxi/limo option, which offers a door to door service. The traveler does not have to carry their bags all over the place and often will receive assistance with them on both ends of the tr ip. The days away variable is negative for drive and park, meaning the longer the trip is, the less likely the person is to choose this mode. Long trips result in high parking costs. Travelers wanting a cheaper option would be smart to choose a different m ode besides this one in this situation. Those arriving after 8:00PM are more likely to choose the drive and drop option. This is a time when those who work during the day are free and able to drop them off. Travelers arriving from within the TAZ range are more likely to choose the shared ride option, and less likely to choose the drive and park option. Short, close trips provide more alternative options than the expensive parking. For those outside the TAZ range, the further away they are coming from, the m ore likely they are to choose the rental option. Weekday travelers are more likely to choose transit options, as this is when they are most readily available, and less likely to choose drive and drop option, as the people who would normally drive them are usually working on weekdays. 4.2.2 Resident Non Business Egress This model featured six different modes to be chosen from. These six modes were drive and drop (picked up), drive and park (parked vehicle), rental car, transit, taxi/limo, and shared ride. Si ngle male travelers are less likely to choose the shared ride option, where single female travelers are less likely to choose the rental option. Parties of two or more are like to choose the drive and park option. Low income travelers are more
45 likely to ch oose the transit option, where high income travelers are more likely to choose the taxi/limo option. The more checked bags a traveler is returning with, the less likely they are to choose the transit option. Dealing with lots of luggage on buses and trains can be a big hassle. Travelers returning from longer trips are less likely to choose the drive and park option, as longer trips result in higher parking costs. Instead, they are more likely to choose the transit option. Those who departed the airport afte r 8:00PM were more likely to be picked up. This is a time when those who work during the day are available to pick people up with no other conflicts. 4.2.3 Resident Non Business Model Comparison The access and egress models for the resident non business market segment were the same and different in many aspects. Some of the variables were only applied to the access model due to data limitations, so these do not have any effect on the egress mode. The access model showed no effects of single trave lers, whether male or female, while the egress model showed males tend to avoid shared ride and females tend to avoid rental cars. For both models, parties of two or more are more likely to choose the drive and park option. Those with children are also le ss likely to choose the transit option for both directions of travel. Low income travelers are more likely to choose the transit option for both access and egress modes. These travelers are looking for a cheaper option, and transit is the cheapest availabl e mode. The length of the trip shows similar effects on both modes, with those on longer trips less likely to choose the drive and park mode. For the egress, those on longer trips are more likely to choose the transit option, an effect not seen in the acce ss model. For both arriving and departing travelers, doing so after 8:00PM makes them more likely to choose the drive and drop option.
46 Table 4 5 Resident non business access model Variable Drive Drop Drive Park Rental Transit Taxi Limo Shared Ride Single Male Single Female 2+ Adults .484 (5.448) .62 (4.698) 3+ Adults .258 ( 1.736) 4+ Adults 2.17 ( 2.147) Children .601 ( 4.602) Low Income .488 (3.813) Medium Income High Income .423 (4.565) .477 (3.417) Missing Income Trips in Last Year .038 ( 4.373) .018 (2.112) International .962 (2.982) Domestic Bags .259 ( 3.063) .178 (2.141) Days Away .279 ( 12.352) OAK .463 (4.74) 2.04 (9.297) .997 ( 5.04) 1.15 (3.427) SFO 1.43 (6.683) 1.89 (6.177) SJC Arr 0:00 8:00 Arr 8:00 20:00 Arr 20:00 24:00 .348 (2.231) Dep 00:00 8:00 Dep 8:00 20:00 .189 ( 2.335) Dep 20:00 24:00 Own Home Origin 1.03 ( 3.124) 1.005 ( 8.146) Other Origin Inside TAZ 1.009 ( 6.886) 1.55 (2.161) Inside TAZ DA Time .011 ( 7.208) Inside TAZ TR Time Outside TAZ Distance .019 (5.598) Weekend Weekday .276 ( 3.019) .408 (2.675) 2001 .326 (3.825) .961 (3.028) .323 (2.458) 2002 Constant .122 (.604) 4.46 ( 12.319) 3.05 ( 11.151) 3.11 ( 16.565) 5.82 ( 7.604) Sample Size 3353 LL Convergence 4196.33 LL Constants 4669.4921 Adj Rsq 0.09929
47 Table 4 6. Resident non business egress model Variable Drive Drop Drive Park Rental Transit Taxi Limo Shared Ride Single Male .439 ( 2.036) Single Female .857 ( 1.715) 2+ Adults .497 (6.154) 3+ Adults .761 (1.724) 4+ Adults Children .635 ( 4.604) Low Income .524 (4.025) Medium Income High Income .524 (5.796) .623 (4.876) Missing Income Trips in Last Year .034 ( 4.083) International Domestic Bags Per Person .32 ( 3.685) Days Away .259 ( 11.623) .549 (4.381) .078 (2.336) OAK .669 (6.017) 1.94 (9.072) .762 ( 3.859) 1.48 (3.877) SFO .318 (2.896) 1.44 (6.685) .31 (2.3) 2.31 (6.561) SJC Arr 0:00 8:00 Arr 8:00 20:00 .274 ( 3.626) Arr 20:00 24:00 Dep 0:00 8:00 Dep 8:00 20:00 Dep 20:00 24:00 .15 (1.864) 2001 .313 (3.819) .659 (1.871) .329 (1.975) 2002 Constant .34 ( 2.298) 4.19 ( 14.3) 2.99 ( 13.63) 2.18 ( 10.398) 4.18 ( 11.979) Sample Size 3353 LL Convergence 4287.525 LL Constants 4669.0633 Adj Rsq 0.08002
48 4.3 Non Resident Business The final sample for the non resident business market segment included 1387 cases. This was the smallest analysis sample of the four market segments. The full models for both the access and egress modes are shown in table 4 8 and 4 9 respectively. Table 4 7 shown below is a cross tabulation o f the access mode and egress mode chosen sample shares. Table 4 7. Non resident business access and egress mode cross tabulation Egress Mode Picked Up Rental Transit Taxi/Limo Shared Ride Drive Drop 9.01 0.79 1.15 1.87 0.50 Rental 4.04 43.62 1.51 1.87 1.37 Access Mode Transit 1.01 0.14 3.75 0.79 1.01 Taxi/Limo 2.31 0.43 0.58 12.55 1.30 Hotel Shuttle 1.30 0.65 0.65 1.73 0.79 Shared Ride 0.72 0.29 0.36 1.51 2.38 Non resident business travelers choose the rental option more than any other option for both the access and egress trip. Nearly thirty percent of travelers in this group choose different access and egress modes. Part of this was the travelers who chose hotel shuttle n eeding to choose another mode. 4.3.1 Non Resident Business Access This model consisted of six modes available to choose from. These six modes were drive and drop, rental vehicle, transit, taxi/limo, hotel shuttle, and shared ride. Single female travelers a re more likely to choose the shared ride option. Not having to travel by themselves gives them a sense of safety. Travel parties with two or more adults are less likely to choose the drive and drop mode. For non resident business travelers, they often do n ot have somebody available to drive them to the airport,
49 especially in large quantities. Low income travelers tend to choose the transit option over all others. Because they have little to spend, they tend to choose the cheapest option available to them. I f the costs of the traveler are going to be reimbursed, they tend to choose the taxi/limo option. This mode generally tends to be the most expensive, but also offers the most flexibility. If you are not worried about cost, this mode will get you from your origin to your destination the quickest and most reliably. Travelers arriving before 8:00AM are more likely to choose the hotel shuttle option. Standard checkout time for a hotel is usually around 11:00AM, so those arriving left a little bit early and the time frame for arrivals to the airport seems correct. Travelers whose origin was a place of business are more likely to choose the rental option, while those with a hotel origin are more likely to choose the hotel shuttle option. 4.3.2 Non Resident Busine ss Egress This model consisted of five possible modes available to be chosen from. These five modes were drive and drop (picked up), rental car, transit, taxi/limo, and shared ride. Single male travelers prefer not to use the taxi/limo option, while single female travelers prefer not to use the rental car option. Low income travelers prefer to be picked up from the airport, which provides ease. High income travelers prefer not to use the shared ride mode option. If the costs of the traveler are reimbursed t hey tend to choose the rental option. For a business traveler, the rental option provides the most flexibility for their trip. Those departing after 8:00PM are more likely to be picked up. The people who would pick them up are normally working during the d ay and become available during this time period. Variables for international travel, children, and quantity of days here did come back statistically insignificant in this model specification.
50 4.3.3 Non Resident Business Model Comparison The fact that this analysis sample was the smallest, consisting of only 1387 cases, may have an effect on which variables are statistically significant or not. Variables that were significant in other model specifications showed no effect on this group of travelers. There we re still similarities and differences between the access and egress models for this market segment. The length of the trip only affected the access mode choice and had no impact on the egress mode choice. For the access model, single female travelers were more likely to choose the shared ride mode, while for the egress model, this effect was not seen. Instead, single male travelers were less likely to choose the taxi/limo option and single female riders were less likely to choose the rental option. The acce ss model showed low income travelers prefer the transit option, while the egress model showed they prefer to be picked up. The access model showed that if the costs were to be reimbursed the traveler preferred the taxi/limo option; while the egress model s howed the traveler preferred the rental option. For most business trips, if a car is rented for the egress portion of travel, the access mode will correspond. For both models, a departure time after 8:00PM made the traveler more likely to choose the drive and drop option, as this is when those who would pick them up would be more available. For variables that were only applied to the access model due to data limitations, only the origin type and drive alone distance for those inside the TAZ range affected w hich mode a person chose. Transit time, distance for those outside the TAZ range, and the day of the week showed no effects in this model specification.
51 Table 4 8. Non resident business access model Variable Drive Drop Rental Transit Taxi Limo Hotel Shuttle Shared Ride Single Male Single Female .870 (3.419) 2+ Adults .487 ( 2.394) 3+ Adults 4+ Adults Children Low Income .595 (2.833) 1.07 (4.170) Medium Income High Income Missing Income Costs Covered 1.11 (7.989) 1.40 (7.309) Costs Not Covered Trips in Last Year International Domestic Bags .495 (3.638) Days Away .094 (2.977) OAK .648 ( 3.719) 1.49 (3.701) SFO 1.08 ( 6.267) 1.44 (3.743) .903 (4.931) SJC Arr 0:00 8:00 .571 (2.173) Arr 8:00 20:00 Arr 20:00 24:00 Dep 00:00 8:00 Dep 8:00 20:00 Dep 20:00 24:00 .566 (1.888) Place of Business Origin .489 (3.56) Hotel Origin 2.53 (4.817) Other Origin Inside TAZ .698 ( 2.675) 1.4 (2.255) 1.23 (2.932) 1.36 (1.805) Inside TAZ DA Time .021 (6.783) Inside TAZ TR Time Outside TAZ Distance Weekend Weekday 2001 .569 (2.566) 2002 Constant .861 (3.028) 3.602 ( 5.1) 2.28 ( 5.099) 4.36 ( 4.95) 1.56 ( 7.154) Sample Size 1387 LL Convergence 1709.844 LL Constants 1939.1027 Adj Rsq 0.11465
52 Table 4 9. Non resident business egress model Variable Drive Drop Rental Transit Taxi Limo Shared Ride Single Male .279 ( 1.931) Single Female .573 ( 3.913) 2+ Adults 3+ Adults 4+ Adults Children Low Income .784 (4.277) .667 (2.618) Medium Income High Income .565 ( 2.234) Missing Income Costs Covered .877 (6.532) .788 (4.71) Costs Not Covered Trips in Last Year .048 ( 2.095) International Domestic Bags Per Person .325 ( 3.161) Days Away OAK .499 ( 2.723) .713 (2.218) .491 ( 1.921) SFO .78 ( 4.357) .769 (2.528) .747 (3.523) 1.01 (4.056) SJC Arr 0:00 8:00 Arr 8:00 20:00 Arr 20:00 24:00 Dep 0:00 8:00 Dep 8:00 20:00 Dep 20:00 24:00 .333 (1.917) 2001 .557 (3.264) .847 (3.532) .597 (3.074) .506 (1.983) 2002 Constant .915 (5.499) 1.61 ( 5.751) .344 ( 1.462) 1.07 ( 4.53) Sample Size 1387 LL Convergence 1775.157 LL Constants 1923.2968 Adj Rsq 0.07285
53 4.4 Non Resident Non Business The final sample for the non resident non business market segment included 2533 cases. The full models for both the access and egress modes are shown in Table 4 11 and 4 12 respectively. Table 4 10 shown below is a cross tabulation of the access mode and e gress mode chosen sample shares. Table 4 10. Non resident n on business access and egress mode cross tabulation Egress Mode Picked Up Rental Transit Taxi/Limo Shared Ride Drive Drop 41.37 1.03 1.42 1.46 1.62 Rental 3.83 23.13 1.46 1.34 0.51 Access Mode Transit 1.70 0.32 3.83 0.63 0.47 Taxi/Limo 1.26 0.39 0.39 3.28 0.71 Hotel Shuttle 0.87 0.95 0.59 0.75 0.47 Shared Ride 1.14 0.28 0.36 1.22 3.20 Over twenty five percent of travelers in this group choose different access and egress modes. Of these, the largest portions are picked up and use a rental car to access the airport for their return trip. 4.4.1 Non Resident Non Business Access This model consisted of six possible mode alternatives to be chosen from. The six mode options were drive and drop, rental car, transit, taxi/limo, hotel shuttle, and shared ride. Single male travelers are less likely to choose the shared ride option, while single female travelers are more likely to choose the drive and drop option. Travelers in parties of two or more are also less likely to choose the shared ride option as well. People traveling with children are less likely to choose the taxi/limo option. Lower income travelers are more likely to choose the transit option, while high income travelers are more likely to choose the taxi/limo mode option. It would make sense that lower income
54 people would prefer a cheaper mode while higher income people would prefer a more expensive mode. Travelers departing to international destinations are more likely to choose the hotel shuttle mode and less likely to choose the rental car mode. When in a foreign country, driving can be very difficult and confusing. Returnin g travelers not choosing this mode makes sense. Also, when visiting another country people most often do not relate to anybody and must stay in a hotel. Using the hotel shuttle to the airport is the easiest access mode for these people. The longer the trip is, the more likely the person is to choose the drive and drop mode. Extended trips to the region are often to visit family or friends. When the trip is over, it is convenient to have these people take you to the airport as you return home. Travelers arri ving between midnight and 8:00AM tend to choose the hotel shuttle option. For non residents, most of them are staying in a hotel. With check out time in the morning, this seems intuitively correct. For e more likely to choose the drive and drop option, while those coming from a hotel are more likely to choose the hotel shuttle option. Those travelers arriving from within the TAZ range are less likely to choose the taxi/limo option. For the shorter trips this represents, there are more modes available and they can choose a different one. Also for those inside the TAZ range, the longer the driving time, the less likely they are to choose the drive and drop mode. The travel time for this mode effectively dou bles, so the longer a trip is, the less likely somebody would want to choose it. Weekday travelers are more likely to choose a transit alternative. Transit runs at peak schedules during the weekdays so this seems correct.
55 4.4.2 Non Resident Non Business E gress This model consisted of five possible more alternatives to be chosen from. These five mode options were drive and drop (picked up), rental car, transit, taxi/limo, and shared ride. Single male travelers are less likely to choose the taxi/limo option, while single female travelers are more likely to be picked up. Lower income travelers are also more likely to choose the transit option because it saves them money. High income travelers are more likely to choose the rental car option. Having a rental car for the duration of the trip makes it very easy to get around. If you have a high income, spending the extra money for this convenience is a smart decision. Travelers arriving from international destinations are less likely to choose the rental car option This is due to the fact there are in a foreign country and the driving laws might be completely different from their home country. The more checked bags a person is traveling with, the more likely they are to choose the taxi/limo option. The longer the stay, the more likely the person is to choose the drive and drop option. Long stays from non residents tend to be while visiting family members or friends. These provide both a place to stay and reliable transportation. At the end of the trip, the person t hey are staying with simply drives them back to the airport for their return flight. Travelers arriving or departing after 8:00PM are more likely to choose the drive and drop option. This is the time of day when the people who would pick them up or drop th em off are available to do so. 4.4.3 Non Resident Non Business Model Comparison The access and egress models for the non resident non business market segment have some similarities and differences. Single males are less likely to choose shared ride for the access trip and less likely to choose taxi/limo for the egress trip, while single female travelers are more likely to choose the drive and drop mode for both directions.
56 Those with children are less likely to choose the taxi/limo mode for the access trip while this variable shows no effects on the egress mode choice. Low income travelers are more likely to choose a transit option for both directions of travel, while high income travelers are more likely to choose different options for both directions of travel. International travelers are less likely to choose a rental car for both directions of travel, and the access only they are more likely to choose the hotel shuttle option. Driving in a foreign country can be a daunting task, and most would not want to approach this. The longer the trip, the more likely a person is to choose the drive and drop mode alternative for both directions of travel. Most of these are likely to people visiting family and friends who they can depend on to drive them both to and from the airport when they arrive and depart. For variables only applied to the access mode, the origin type, driving time from the origin, and the day of the week all showed significant effects on mode choice. Variables that were not significant in either the access or egress model included large parties and the number of trips from the airport in the last year. Most non residents would not take multiple trips to a region they do not live in the past year, so this variable showing no effects seems correct.
57 Table 4 11. Non resident non business access model Variable Drive Drop Rental Transit Taxi Limo Hotel Shuttle Shared Ride Single Male .882 ( 6.684) 1.04 ( 3.61) 1.17 ( 4.52) Single Female 1.36 (10.653) .707 (3.497) 2+ Adults .537 ( 2.371) 1.23 ( 5.699) 3+ Adults 4+ Adults Children .469 ( 2.686) Low Income .215 (1.869) .538 (3.035) Medium Income High Income .435 (3.475) .49 (2.301) Missing Income Trips in Last Year International .479 ( 2.443) 1.04 (3.353) Domestic Bags .145 (2.108) Days Away .095 (3.455) OAK 1.84 (5.931) .84 ( 2.547) 2.25 (3.638) SFO .402 (3.486) 1.58 (5.233) .836 (3.606) .923 (3.391) 3.62 (6.101) SJC Arr 0:00 8:00 1.16 (4.937) Arr 8:00 20:00 Arr 20:00 24:00 .746 (3.841) Dep 00:00 8:00 Dep 8:00 20:00 Dep 20:00 24:00 Somebodies House Origin 2.49 (15.610) .567 (3.548) .463 (2.173) Hotel Origin 3.08 (5.152) Other Origin Inside TAZ .97 ( 6.52) 1.36 ( 2.52) 1.04 ( 4.2) Inside TAZ DA Time .009 ( 4.791) Inside TAZ TR Time Outside TAZ Distance .049 ( 2.265) Weekend Weekday .393 (2.181) 2001 .288 (2.928) 2002 Constant 1.85 (9.243) 2.09 ( 6.041) 1.47 (2.352) 4.19 ( 6.668) 1.18 ( 1.849) Sample Size 2533 LL Convergence 2762.031 LL Constants 3456.1692 Adj Rsq 0.19818
58 Table 4 12. Non resident non business egress model Variable Drive Drop Rental Transit Taxi Limo Shared Ride Single Male .944 ( 4.435) Single Female .703 (6.543) .476 (2.552) 2+ Adults .738 (6.969) 3+ Adults 4+ Adults Children Low Income .469 (4.272) .592 (3.418) .564 (3.064) Medium Income High Income .377 (3.196) Missing Income Trips in Last Year International .511 ( 2.606) Domestic Bags Per Person .159 (2.25) .208 (1.999) Days Away .129 (5.389) OAK 1.88 (6.881) 1.00 (2.815) SFO .277 (2.596) 1.68 (6.216) 1.31 (8.017) 2.38 (7.729) SJC Arr 0:00 8:00 Arr 8:00 20:00 Arr 20:00 24:00 .670 (3.947) Dep 0:00 8:00 Dep 8:00 20:00 Dep 20:00 24:00 .543 (5.062) 2001 .399 (3.986) .374 (2.434) 2002 Constant .461 ( 2.983) 2.51 ( 8.9) 1.42 ( 7.591) 2.87 ( 8.89) Sample Size 2533 LL Convergence 2999.510 LL Constants 3286.9332 Adj Rsq 0.08510
59 CHAPTER 5 SUMMARY AND CONCLUSI ONS The field of mode choice modeling for airport access trips is still an evolving field. The large volume these trips represent lends itself to making the field more prominent within transportation demand modeling in the future. The fact that trips are curre ntly either ignored in models or are accounted for separately is not a satisfactory solution to this problem. With the addition of these models into standard travel demand models in the future, the planning applications and benefits from the airport side w ill become increasingly obvious. With air traffic projected to steadily increase well into the future, the magnitude these trips represent will continue to rise as well. Data used in this study came from the 2001 and 2002 Air Passenger Survey conducted by the MTC in California. Eight models were formulated using a Multinomial Logit Model Structure. The eight models were separated in four access and four egress models for each of the four main market segments. The four market segments were resident business, resident non business, non resident business, and non resident non business. 5.1 Model Discussion In the past, the explanatory variables related to travel composition have been very limited in the field. This study set out to bridge this gap and look at s ome variables that had not been considered in the past. While every model previously studied included travel time and travel cost variables, it was important to note that a statistically significant model can be formulated without these two variables, howe ver implicit effects of these two can still be seen in the final models. The travel time was applied only to the access model due to data limitations. The fact that higher income individuals generally prefer more expensive modes and lower income individual s generally prefer cheaper
60 modes shows the same effects as the travel cost variable in past studies. For business travelers, it is important to note that both income and whether or not the costs of the chosen mode would be reimbursed have an effect on whic h mode is chosen. Where applicable, the signs and interpretations offered by explanatory variables from past studies did match up. The most commonly used variable in the past was household income, which showed higher incomes chose more expensive modes. Thi s was confirmed again in the models formulated for this study. In addition, the more checked bags a person was traveling with, the more likely they are to choose a door to door service mode. This prevented them from carrying luggage on transit options. Thi s was again supported in the models formulated for this study. For effects only shown on the access models, the origin location type and travel time affected the mode a person chose. The day of the week being a weekday or weekend would have an impact on t he mode choice as well. The second part of the study was to create egress mode choice models, which had not previously been done. Using the same data as the access models, the egress models were successfully formulated with all variables being statisticall y significant. One of the main reasons for the egress models not being formulated in the past was due to data limitations. As is true with any transportation survey, incomplete responses were prevalent in the dataset obtained. Working around this limited t he sample size of the final dataset, so it is possible more noticeable effects can be seen with more data. From the cross tabulations shown for the access mode and egress mode, depending upon market segment the variation ranged from fifteen to thirty perce nt choosing different modes. This statistic supports the need for the egress mode to be modeled. If all travelers chose the same mode for access and egress, a
61 singular access mode choice model would be sufficient. Certain assumptions could be made and the model easily converted to an egress mode choice model. Because there is large variation in the modes chosen, a separate model is necessary. While travel surveys would need to be revised to account for the fact the egress mode is just as important, this can be done fairly simply moving forward. 5.2 Conclusions Airport access mode choice models a re not only important to predict what mode travelers choose to get to and from the airport, but play a critical role in the planning process. When airports look at ad ding an additional mode such as an exclusive transit option, the models become very useful. For example, travelers with low income are a very good predictor of which people will choose transit options. With the addition of an exclusive rail line servicing the airport, estimating which of these will stick with the new transit option based on revised cost and time estimates would be important. The models are also important for parking allocation purposes. Using the models to accurately predict which travelers will use a rental car or drive and park can determine the size and capacity of these facilities needed. In addition to be ing used from an airport planning perspective, the models can also be used on a regional level. Travel demand models often do not acco unt for tourist travel or how they are affecting the local network. Using the models to predict the mode these extra travelers use can be important when looking at capacity of the network. For example, knowing how many people rent a car would be very usefu l in this situation. High income is an accurate predictor of which non residents will rent a car so if more high income tourists start arriving, the network may see an increase in activity. Using
62 this information to plan for an influx in extra traffic can be a big advantage on a regional level. One of the biggest needs for these models stems from the fundamental different between access and egress mode choice and the traditional mode choice model of the four step process. In the traditional model, the mode options are very limited. With the airport model, the mode offerings are wide and very different. Taxi/limo options in addition to shared ride options are not available in the traditional model. In addition, the variables included in the models are very d ifferent as well. In the airport model, variables such as number of checked bags and trip destination being domestic or international play an important role. These types of variables would never be thought about in simpler mode choice models, which facilit ate the need for the airport and access mode choice models to be estimated. 5. 3 Future Research With the field of airport access mode choice modeling continuing to grow, one important piece of the puzzle is the travel surveys. Without reliable data, the models themselves cannot be estimated. From proper design to efficient distribution, the survey it self will remain an integral part of the process. With the data in hand, the models themselves will continue to evolve. With the future of transit options and high speed rail in this country, more alternative modes will be available for access to airports. With the end goal of having the best possible model performance, many improvements can be made. For this study, only cases with both the access and egress mode were considered. Removing this filter creates a much larger dataset to be used. Assumptions als o need to be made in order to use the egress location. While destination of the egress trip was not considered in this study, it would be useful to look at affects of
63 different types of locations. The access and egress models themselves will continue to be improved and used to make planning decisions in the future. The underlying difference between traditional travel demand models and the more specific airport access and egress mode choice models is the modes available. This fundamental difference facilitat es the need for these special types of models. Moving forward, it can be seen why both the access and egress mode choices need to be accounted for and successfully implemented into travel demand models.
64 LIST OF REFERENCES Alhussein, Saad N. "Analysis of Gr ound Access Modes Choice King Khaled International Airport, Riyadh, Saudi Arabia." Journal of Transport Geography 19.6 (2011): 1361 67. Print. Bondzio, Lothar. "Study of Airport Choice and Airport Access Mode Choice in Southern Germany." Proc. of European Transport Conference 1996. N.p.: n.p., n.d. N. pag. Print. Choo, Sangho, Ikki Kim, and Soyoung You. "Exploring Charateristics of Airport Access Mode Choice in Korea." TRB 86th Annual Meeting Compendium of Papers CD ROM : January 21 25, 2007, Washington, D.C. Washington, D.C.: Transportation Research Board, 2007. N. pag. Print. Cirillo, Cinzia, and Renting Xu. "Forecasting Cybercar Use for Airport Ground Access: A Case Study at BWI (Baltimore Washington International Airport)." Journal of Urban Planning a nd Development 136.3 (2010): 186 94. Print. "Enplanements at Primary Airports (by Rank)." Passenger Boarding (Enplanement) and All Cargo Data for U.S. Airports N.p., n.d. Web.
65 Choice of Access Mode Transportation Research Part E 48.5 (2012): 1023 31. Print. Mamdoohi, A. R., M. Saffarzadeh, A. Taherpour, and M. Yazdanpanah. "Modeling Air Passengers' Ground Access Mode Choice a Case Study of IKIA." International Journal of Modeling and Optimization 2.2 (2012): 147 52. Print "MTC -Maps and Data." MTC -Maps and Data N.p., 9 July 2007. Web.
66 BIOGRAPHICAL SKETCH Benjamin Reibach, originally from Philadelphia, Pennsylvania, enrolled at the University of Florida in August of 2011.He joined the Transportation Graduate program at UF following completion of his B.S. degree in Civil Engineering from the Pennsylvania State University. In addition to this, he is a licensed Engineer in Training, working towards his Professional Engineer lic ensure. As a graduate research assistant under the direction of Dr. Siva Srinivasan, he has worked on projects including activity based model implementation into FSUTMS and closely studying auxiliary transportation demands. Outside of educational and profe ssional areas, he is an avid sports fan, following every team from his hometown and his former school. Favorite sports of his include football and basketball, with the Eagles and 76ers being his favorite teams. He is also passionate about volunteering to b etter serve the community and has done countless events while a student at UF.