Scholarly article on topic 'Understanding Immigrants’ Mode Choice behavior in Florida: Analysis of Neighborhood Effects and Cultural Assimilation'

Understanding Immigrants’ Mode Choice behavior in Florida: Analysis of Neighborhood Effects and Cultural Assimilation Academic research paper on "Social and economic geography"

CC BY-NC-ND
0
0
Share paper
Academic journal
Transportation Research Procedia
OECD Field of science
Keywords
{Immigrants’ / "travel behaviour" / "mode choice modeling" / "neighbourhood effects" / "principal component analysis"}

Abstract of research paper on Social and economic geography, author of scientific article — Hamidreza Asgari, Nishat Zaman, Xia Jin

Abstract The goal of this study is to develop Multinomial Logit models for mode choice behaviour of immigrants, with key focuses on neighbourhood effects and behavioural assimilation. The first aspect assumes a linkage between social network ties and immigrants’ chosen mode of transportation, while the second dimension explores the gradual propensity towards alternative mode usage with regard to immigrants’ settlement period in the United States. Factor analysis was carried out to establish neighbourhood typologies based on income, family structure, and education. Mode choice models were then developed for work, shopping, social and recreational, and other trip purposes to evaluate the impacts of various land use patterns, neighbourhood typology, and socioeconomic and demographic attributes on immigrants’ travel behaviour. Estimated coefficients for each alternative mode (i.e., high-occupancy vehicle, public transit, and non-motorized transport) were compared to single-occupancy vehicles (SOV). The model results revealed significant influences of neighbourhood and land use variables on the usage of alternative modes among immigrants. Incorporating these indicators into the demand forecasting process is expected to provide a better understanding of diverse travel patterns for the unique composition of population groups in Florida.

Academic research paper on topic "Understanding Immigrants’ Mode Choice behavior in Florida: Analysis of Neighborhood Effects and Cultural Assimilation"

Available online at www.sciencedirect.com

Transportation

Procedía

Transportation Research Procedía 25C (2017) 3083-3099 ■ ■ w «J «J

www.elsevier.com/locate/procedia

World Conference on Transport Research - WCTR 2016 Shanghai. 10-15 July 2016

Understanding Immigrants' Mode Choice behavior in Florida: Analysis of Neighborhood Effects and Cultural Assimilation

Hamidreza Asgaria, Nishat Zamanb, Xia Jinc

"PhD, Department of Civil and Environmental Engineering, Florida International University, Miami, FL 331 74, USA bM.Sc, Department of Civil and Environmental Engineering, Florida International University, Miami, FL 331 74, USA cPhD, AICP, Assistant professor, Department of Civil and Environmental Engineering, Florida International University, Miami, FL 33174, USA

Abstract

The goal of this study is to develop Multinomial Logit models for mode choice behaviour of immigrants, with key focuses on neighbourhood effects and behavioural assimilation. The first aspect assumes a linkage between social network ties and immigrants' chosen mode of transportation, while the second dimension explores the gradual propensity towards alternative mode usage with regard to immigrants' settlement period in the United States. Factor analysis was carried out to establish neighbourhood typologies based on income, family structure, and education. Mode choice models were then developed for work, shopping, social and recreational, and other trip purposes to evaluate the impacts of various land use patterns, neighbourhood typology, and socioeconomic and demographic attributes on immigrants' travel behaviour. Estimated coefficients for each alternative mode (i.e., high-occupancy vehicle, public transit, and non-motorized transport) were compared to single-occupancy vehicles (SOV). The model results revealed significant influences of neighbourhood and land use variables on the usage of alternative modes among immigrants. Incorporating these indicators into the demand forecasting process is expected to provide a better understanding of diverse travel patterns for the unique composition of population groups in Florida.

© 2017 The Authors. PuMished by Elsevier B.V.

Peer-review imder responsibility of WORRLD CONFERENCE ON TRANSPORT RRESEARCH SOCIETY.

Keywords: : Jmrngrante' travel behaviour, mode ctoce mo^hn^ neighbourhood effect^ principal component analysis

L Introduction

The Umted States is consMered as a popukr hosting countiy for mm^rantts tiesMes AustraHa and Canada [Massey et aL, 1993]. Immgranits and theh cMMren and grandchildren have accounted for half of the tote1 mcreased population over the last two decades in the US [Passe1 and Cohn, 2009]. Specifically, immigrant population stood 13% of the

2352-1465 © 2017 The Authors. Published by Elsevier B.V.

Peer-review under responsibility of WORLD CONFERENCE ON TRANSPORT RESEARCH SOCIETY. 10.1016/j.trpro.2017.05.319

total US population; and accounted for 16.4% of the workers engaged in the US civilian labour force in 2012 [Martin and Midgley, 2010]. Given the significant contributions of immigrants to the demographic, social, economic and cultural aspects of the country, many researchers have studied immigrants' housing, health, education, and employment conditions; while the transportation aspects for immigrants especially their attitude towards different travel modes and choices haven't received as much attention.

Research efforts to-date suggest that immigrants reveal an aberrant travel behaviour compared to US- born residents and that they have higher usage of transit, carpool, and non-motorized modes [Blumenberg, 2009; Burbidge, 2012; Casas et al., 2004; Handy et al., 2008]. From this perspective, it is critical for planning agencies to understand their travel behaviour for today's transportation decisions, and even more to understand the implications in terms of the influence of the following generations in shaping the future travel pattern to promote sustainable and liveable transportation. However, much remains unexplored on the underlying factors, and the transferability of their impacts, that lead to the higher usage of alternative modes among immigrants. These factors may involve cultural preferences related to social network and life style, such as large households, close coordination among household members and friends, or living in dense, mixed-use neighbourhoods. Other factors may be more passive, such as lower car ownership and legal barriers to driving, etc.

Taking the above into account, the objective of the current study is to investigate the mode choice behaviour of immigrants, focusing on neighbourhood effects and behavioural assimilation. The research questions to be answered through this study are:

• Controlling all other factors (including density, size, and other land use variables), does living in an immigrant neighbourhood lead to higher usage of alternative modes?

• Is the level of preference towards transit and non-motorized modes associated with the number of years lived in this country, controlling for other factors (such as income, household lifecycle, auto ownership, etc.

This study will provide additional insights on the underlying factors that contribute to the mode choice behaviour of immigrants beyond the factors that are known to have significant influence on the use of alternative modes, such as land use variables, socio-economic and demographic characteristics. The study results could provide better tools to facilitate policy and investment decisions to meet the travel needs for immigrant communities and inform the efforts in promoting sustainable transportation. In addition, incorporating immigrant indicators into the demand forecasting process holds the promise in enhancing the performance of the models to better reflect the diverse travel patterns of the market.

2. Literature review

Travel patterns tend to vary with respect to individual, cultural and geographical characteristics [Eisenhauer et al, 2007]. Not only do people from different occupations, attitudes, and lifestyles possess different types of travel behaviour [Crane and Crepeau, 1998], but also travel patterns could be recognized as indicators of their preferences, habits, availability of opportunities and the constraints they face on a short or long-term basis [Blumenberg and Smart, 2010].

Many studies revealed that immigrants are more inclined toward the use of alternative modes, i.e., carpooling, ride sharing, public transit, walking, and biking, than US-born individuals [Crane and Crepeau, 1998; bagley and Mokhtarian, 2002; Blumenberg, 2008; Liu and painter, 2012]. Such behaviour is affected by a number of underlying factors, including various opportunities that allow immigrants to survive with the use of alternative travel modes or certain constraints that make private vehicle ownership unattainable.

The literature reveals a direct relationship between survival opportunities and residential neighbourhood. Accordingly, immigrants tend to live in neighbourhoods with higher residential and employment densities where destinations are reachable by shorter walking or biking trips [Bagley and Mokhtarian, 2002; Blumenberg, 2008; Smart, 2010], or in dense urban neighbourhoods with extensive transit service [Liu and Painter, 2012]. One study showed that immigrants reside in metropolitan areas at a rate of twice the US-born population [Crane and Crepeau, 1998]. A deeper investigation of their settlement process showed that during their migration stage, immigrants had a tendency

to reside in inner city enclaves of disadvantaged areas where the majority of people with limited socio-economic resources reside [Blumenberg, 2008]. Living in the same or nearby ethnic neighbourhood made it favourable for immigrants to choose alternative transportation, such as carpooling, car-borrowing, or ride-sharing, probably because it is easier for them to maintain social networks [Alba et al., 2000]. The existence of such social ties was demonstrated earlier [bagley and Mokhtarian, 2002]. Factors like spatial proximity and time schedule compatibility play significant roles in the likelihood of receiving transportation support [Blumenberg and Smart, 2009]. Immigrants also face linguistic and driving license constraints, higher auto insurance rates, and the financial inability to afford a private vehicle at early stages of migration to the US [Smart, 2010; Smart, 2014]. Such constraints provide further motivations for immigrants to rely on alternate options such as carpooling and ride sharing.

Immigrants' travel behaviour may also be influenced by their gradual assimilation to the American culture. Immigrants gain the economic similarity of middle-income Americans after living more than thirty years in the US [Alba, 2000]. Immigrants' household characteristics, both from geographic and demographic aspects, change over their migration period and impact their travel patterns. In the case of geographic dimension, immigrants tend to navigate within better neighbourhoods of residential amenities. They experience social mobility and become acculturated in the mainstreaming lifestyles of US-born Americans [Blumenberg and Smart, 2014]. Moreover, the development of their economic condition broadens their outlook [Lovejoy and Handy, 2011]. In demographic features, most of the newer immigrants have smaller families and show fewer propensities to drive alone than those who have been in the US for more than 10 years [Smart, 2010; Tal and Handy, 2010]. However, as time passes by, their household sizes increase with the birth of children, necessitating more trips and moving forward to owning private vehicles [Smart, 2010].

Individuals' household, personal, and travel information were derived from 2009 National Household Travel Survey (NHTS) Florida add-on data [National Household Travel Survey (NHTS), 2009; NHTS 2009 user guide, 2009]. Close to 4,000 immigrants and 14,000 trips associated with them were identified from the data set for the purpose of this study.

Tract level SED information was obtained from the 2010 American Community Survey (ACS) data [US Census Bureau, 2010], based on which, neighbourhood typologies and location quotient of immigrant concentration were derived. Employment information was obtained from the InfoUSA data. The land use data at the tract level were derived from The Smart Location Database, a free data product and service provided by Smart Growth Program of the US Environmental Protection Agency (EPA) [Smart Location Database, 2014].

3.1. Land use and built environment variables

Built environment can affect travellers' behaviors [Sharifi et al., 2014, 2015, 2016]. In this study, the built environment pattern is investigated through the land use indices. Data from 2010 American Community Survey (ACS), InfoUSAgroup and the US EPA Smart Location Database are compiled to derive various land use variables at the census tract level using GIS spatial analysis tools. The built environment indices include entropy index, job housing balance index, measure of attractiveness, population and job density, employment entropy index and auto accessibility. A list of BE variables along with formulas and descriptions is presented in table 1.

Table 1 Built Environmental and Land-use Variables

BE Variables Theoretical Formula Description

3. Data

A"j =Area of land use i in tract j D# =Area of tract j

N# =Number of different land uses in tract j

It quantifies the level

Entropy Index

types within a certain geographical area.

of balance among different land use

Table 1

Built Environmental and Land-use Variables (continued)

Job-Housing Balance Index

Measures of Attractiveness

Population Density Job Density

Employment Entropy Index

Auto Accessibility

Abs(T.E. -1.5 X T.H.U.) T.E.= Total employment

T .E. +1.5 X T .H U.

T.H.U.=Total housing units

TEC# = Total number of employment centers in tract j A# = Area of tract j

Popj = Population in tract j A# = Area of tract j

NJ# = Number of jobs in tract j A# = Area of tract j

^ x ln( ^ )

NTEj TE/

EmPj x f(dhj

f (d= a x dj x dj'

E5"j = Employment type i within a 5-tier classification scheme

TEj = Total employment in tract j

Nj= Number of employment types in tract j

Empj = measure of working-age population in the CBG j f(d)ij = measure of impedance between CBGs i and j

It indicates whether there is an adequate supply of housing to house workers employed in a defined area, or vice versa. It explores the disribution of job/employment centers within each census tract.

Population per square mile

Population per square mile

It is used to show the land use diversity based on employment type.

It measures jobs within a 45-minute commute via automobile from residence.

4. Methodology

Neighbourhood effects and behavioural assimilation are the two major focuses of this research. While built environmental variables have been widely used in travel behaviour studies, this study steps further and tries to view the concept of "neighbourhood" beyond what's being generally considered as built environment factors or land use indices. In view of that, it is assumed that a neighbourhood could be defined as a cluster of people who share similar socio-economic and demographic attributes as well as spatial proximity.

Two major properties are considered in measuring neighbourhood effects: immigrant concentration, which calculates the ratio of immigrants to native US born inhabitants in each census tract; and neighbourhood typologies, which are used to delineate different types of neighbourhoods with respect to common socio-economic and demographic attributes among the residents.

The concentration of immigrant population in a specific tract is defined by the location quotient (LQ). Based on the threshold values of a location quotient, a census tract can be regarded as either an immigrant neighbourhood or a non-immigrant neighbourhood [Eisenhauer et al., 2007]. The location quotient of immigrant/foreign-born population concentration of a certain census tract i is defined as

LQi =■

et and e are the foreign-born and total population, respectively, in census tract i. E" and E are the foreign-born and total population, respectively, in the State of Florida.

Based on the outcome values, which range from 0.0 to 4.6, the tracts are grouped into five different categories: 1) Concentration less than the State, 2) Equal to the State, 3) Up to two time the State, 4) Up to three time the State, and 5) Up to four times or more than the State. The straightforwardness of the equation suggests that each tract LQ is independently calculated and not biased by one another.

Table 2.

Tracts Based on Location Quotient Value of Immigrant Concentration.

LQ Value Immigrant Concentration Level of Tracts Number of Tracts

LQ <1.0 Less than the state 851

LQ =1.0 Equal to the state 71

1.0< LQ < 2.0 Up to two times the state 434

2.0< LQ < 3.0 Up to three times the state 125

3.0< LQ< 4.6 Up to four times or more than the state 45

For the purpose of neighbourhood delineation, variables were selected following the existing literature on neighbourhood definitions in order to reflect the very basic social background of inhabitants. In order to treat scale issues, variables were converted into percentages for each census tract. The variables used in this study include: education level, income, and household structure (with children aged 18 years or less, divorced/separated/widowed, or single living). Factor analysis was used to identify the latent neighbourhood dimensions based on these variables using Principle Component Analysis (PCA) as the major extraction approach. The Communalities value revealed that at least 65% of each variable's variance has been reproduced by the extracted factors whereas 30% is the recommended criteria. The Eigenvalue suggested that 81% of the total variances have been explained by the factor loadings whereas the recommended value is 75%. Based on the factor loadings on the above variables, two potential factors were recognized and labeled as "Social Status" and "Family Bonding".

Table 3.

Rotated Factor Loading Pattern of the Neighbourhood Social Characteristics.

Component

Neighbourhood Social Status

Neighbourhood Family Bonding_

Percent with less than high school degree Percent with more than college degree Percent with income less than $25,000/Year

Percent with income more than or equal to $50,000/Year

Percent of families with dependent children under 18 years old

Percent of divorced/ widowed/ separated Percent of non-family living alone

0.923 -0.936 0.799

-0.834

0.360 0.091

-0.123 0.029 0.422

-0.436

-0.862

0.727 0.926

A cluster analysis was further carried on the sample based on the two derived factors which results in the segmentation of sample dataset into three mutually exclusive neighbourhood groups: 1) Low Social Status along with High Family Bonding (532 tracts), 2) High Social Status along with High Family Bonding (483 tracts), and 3) High Social Status with Low Family Bonding (487 tracts).

Table 4._Distribution of Census Tracts under Two Neighbourhood Factors.

Factor Neighbourhood Typologies

Type 1 (N=532) Type 2 Type 3 (N=483) (N=487) Outliers (N=15)

1 : Neighbourhood Social Status Structure a 2: Neighbourhood Family Bonding Structure b 0.95 -0.27 -0.82 -0.3 -0.75 1.02 2.01 0.67

a Higher scores represent lower level of educational attainment and household income.

b High scores represent more broken family/ single living households than family living with dependent children in the household.

5. Model structure

The application of discrete choice models has been well established in travel behaviour studies [Vaziri et al. 2014, Shams et al. 2015, Asgari et al. 2014, Asgari & Jin 2015; Sobh, 2015, Khalilikhah et al. 2016], especially when it comes to mode choice analysis [Ben-Akiva and Lerman, 1985; Train, 2003; Cascetta, 2009; Washington et al., 2011; Soltani-Sobh et al., 2016].

At this stage, four Multinomial Logit models have been developed for work, shopping, and social/recreational purposes to examine the influence of the determinant factors including personal, household, land use, neighbourhood, and immigrant attributes on their mode choice behaviours. Significance of immigrant indicators was examined by incorporating additional neighbourhood and assimilation variables. The model takes the functional form of equation

15, where pmo7e is the probability that a certain mode is chosen, Pmode is the Odds ratio (OR) that indicates the

>~Pmode

ratio of the probability of choosing a certain mode over the probability of not choosing the mode, ln Pmode is the

>~Pmode

log odds, PR is vector of person-level variables that has been proved to be a determinant in mode choice behaviour in previous literature, HH is the vector of household-level determinants of mode choice, LU is the vector of land use variables, as well as built environment determinants of mode choice, LQ is the vector of the location quotient of immigrant population concentration, IM is the vector of immigrant status-related variables, NH is the vector of neighbourhood typologies, and a and are estimated coefficients using the maximum likelihood estimation.

Logit (p) = In = a + b PR + b HH + fc LU + b4 LQ + b5 IM + b6 NH (2)

Alternatives include single-occupant vehicle (SOV) representing drive alone (the base category), carpool or shared ride (i.e., all auto trips with more than one occupant), public transportation, and walking/biking as non-motorized transport (NMT). The personal and household control variables, including gender, educational attainment, driver status, household income group, household type, home ownership, household life cycle, etc., are used as categorical dummy variables, while age, number of household adult members and vehicles, household size, etc., are considered continuous variables. Among the immigrant-related variables, the foreign-born or immigrant status of a trip maker is used as a dummy variable, whereas the earliest arrival period reported by any of the household members is considered the categorization criterion for the foreign-born households and further creation of the dummy variables. In this context, the presence of one or more foreign-born members is used as the indicator for considering a household as foreign-born/immigrant household. The derived land use variables are used as continuous variables; whereas the location concentration neighbourhood typologies were considered categorical dummy variables.

After proper data processing, a total of 53,314 trip records remain for the final model development. Out of these trips, 7,132, 14,371, 20,058, and 11,753 trips are classified as work, other (e.g., school, maintenance, etc.), shopping, social, and recreational trips, respectively. The non-home based trips are not taken into account as they are not assumed

to directly represent the mode choice behaviour describing the trip maker's household location choice, socio-economic background, etc.

6. Results and discussions

The multinomial logistic regression models were developed for work, shopping, and social/recreational trips. Attributes of personal, household, immigrant status, land use, and neighbourhood were considered independent variables. The initial assessment of the model was done through the Goodness-of-fit test measures and the statistical test. The Rho-square values for work, shopping, and social/recreational, are reported as 0.719, 0.450, and 0.223 respectively. Previous studies on MNL models have shown similar results [Enam and Chodhury, 2011; Saha, 2010].Specifically, for work trip model, the value exceeded 0.70, which indicates that the predicted models are close to perfect. The detailed results of the multi-nominal logistic regression model of each trip are shown in Tables 5 through 7.

6.1. Individual/household attributes

As expected, the models bode for significant contributions of individual/household characteristics to mode choice decisions. In most cases, results comply with previous findings in literature. Accordingly, older individuals, males, white ethnicity, licensed drivers, and high income households are more likely to drive alone. Significant impacts of household structure and housing type are also observed.

6.2. Immigrant-related attributes

Previous research supports that foreign-born people and immigrants are more likely to use alternate modes of transportation for all trip purposes (carpooling/ride sharing, public transit, and non-motorized modes) than US-born individuals [Batalova, 2014; Blumenberg, 2009; Burbidge, 2012; Casas et al., 2004; Crane and Crepeau, 2004; Bagley and Mokhtarian, 2002; Blumenberg, 2008; Smart, 2010; Tal and Handy, 2010; Liu and Painter, 2012]. For work and shopping trips, they are less likely to carpool than use public transit and NMT. It may be because obtaining a ride from someone else is not always possible, or they live near their place of employment and shopping areas.

Immigrants' arrival period in the US has a significant influence on mode choice. For example, immigrants that migrated 15-20 years ago are more likely to drive alone. In view of work purposes, it is generally observed that the tendency to use alternate modes decreases with number of settlement years in US. This complies with the overall assumption that immigrants grow economically as they spent more years in the US. One exception is for individuals with more than 20 years immigration period, who reflect higher likelihood for HOV usage. A quick overview of this subsample shows that they are either senior people who are incapable of driving alone, or perhaps they have developed large families with insufficient number of private vehicles. For shopping trips, the behaviour is a little more complicated. During their settlement period, not only do individuals grow economically, but also they are more likely to develop social networks, become acquainted with public transport systems, and arrange for joint activities (such as shopping). This may somewhat justify why immigrants with a settlement period of 10-15 years are more likely to use HOV or public transport systems. A similar pattern could be observed for the utility of public transport for social/recreational trips.

6.3. Land use attributes

In terms of work trips, the employment entropy index increases the likelihood of carpooling or ride sharing. It may indicate the fact that a balanced distribution of jobs help people save time and avoid time overlapping to obtain rides. The measure of attractiveness, which may be a sign of higher provision of transit services, has a positive influence on public transit usage. This may indicate that areas that have higher points of interest also hold a higher provision of transit services. For shopping trips, both the entropy index and measure of attractiveness increase the propensity to use public transit and non-motorized modes, which sounds reasonable. It is acceptable to see higher rates of walking/cycling in areas with a mixture of residential and shopping centers nearby. Moreover, the more the land use is mixed with points of interest and equipped with transit services, the more people welcome using alternative modes

of transportation. In terms of recreational trips, the job-housing balance index has a negative impact on NMT. It may indicate that the desire for non-motorized recreational trips is limited to suburban areas or far away from busy city's rhythms.

Table 5._Determinants of Mode Choice Model Outcome for Work Trips.

HOVa Public Transitb NMTc

Intercept (a) 3 29*** -2 31*** 1 9**

PERSONAL ATTRIBUTES

Age -0.008*** -0.017**

Driver Status -5.22*** -2.36*** -5.39***

Education

Less than high school graduate 0.753*** 0.726**

High school graduate, include GED 0 27***

Some college or Associate's degree -0.698** 0.894***

Race of HH respondent

White -1.05***

African American, Black 0.439*** 0.924**

Imputed HH Resident's Race and Ethnicity Combined

Hispanic Non-Black 0.622*

Gender-Male -0.337*** 0.566***

HOUSEHOLD ATTRIBUTES

Household Income Per Year

Income Group 1 (<$5,000-$24,999) 1 47***

Income Group 2 ($25,000-$49,999) 0.436*

Income Group 3 ($50,000-$74,999) 0.343*** -0.81*

Household Hispanic status -1.13**

Type of HH

Detached single house -0.721***

Row house or townhouse 1 11***

HH Size 0.253***

HH Vehicle Numbers -0.715***

HH Life Cycle

One adult, no children -0.744*** 0.861***

2+ adults, youngest child 6-15 yrs 0.385*** 0 971***

Home Ownership (Home is owned) -0.263*

IMMIGRANT RELATED ATTRIBUTES

Respondent is Immigrant/Foreign-born 0.379*** 0.715** 0.552*

Arrival Time

Migrated 20+ yrs ago 0.299**

Migrated 15 to 20yrs ago -1.89***

Migrated 10 to 15yrs ago 0.664**

Table 5._Determinants of Mode Choice Model Outcome for Work Trips (continued)

HOVa Public Transitb NMTc

Migrated 5 to 10 yrs ago 0.962***

Migrated 0 to 5 yrs ago 0.664* 1.788*** 1 453***

LAND USE ATTRIBUTES

Land Use Index Measurements

Measure of Attractiveness 0.299**

5-tier Employment Entropy 1.333**

NEIGHBOURHOOD ATTRIBUTES

Immigrant Concentration (Location Quotient)

Less than state -0.767*** 0.646***

Up to twice the state 0.536**

Up to thrice the state 0.726* 1 27***

Neighbourhood Typologies

Low social status and high family bonding 0.31*** 0.57**

High social status and high family bonding -0.174* -0.787***

High social status and low family bonding -0.267*** -0.575** 0.485**

Initial Assessment Statistics

N (Trips Records) 7132

Null log-likelihood -9887

Cte log-likelihood -3258

Final log-likelihood -2781

Rho-square 0.719

Adjusted rho-square 0.714

a,b,c SOV was taken as the base mode * Statistically significant at the 10% level (i.e., p<0.10); ** Statistically significant at the 5% level (i.e., p<0.05); *** Statistically significant at the 1% level (i.e., p<0.01).

Table 6._Determinants of Mode Choice Model Outcome for Shopping Trips.

HOVa Public Transitb NMTc

Intercept (a) 3.85*** -0.517* 2.06***

PERSONAL ATTRIBUTES

Age -0.014***

Driver Status -3.61*** -3.03*** -3.97***

Education

Some college or Associate's degree -0.13*** -0.328***

Graduate or Professional Degree -0.167***

Race of HH respondent

African American, Black 0.706***

Hispanic/Mexican 0.667***

Imputed HH Resident's Race and Ethnicity Combined

Hispanic Non-Black 0.245*** 0.822***

Non-Hispanic Other Race 0.227***

Gender-Male -0.195*** 0.386***

HOUSEHOLD ATTRIBUTES

Household Income Per Year

Income Group 1 (<$5,000-$24,999) 0.502***

Type of HH

Detached single house -0.3*** -0.369***

Row house or townhouse 0.883*** 0.277**

HH Size 0.047*** 0.646***

HH Vehicle Numbers -2.74*** -0.447***

HH Life Cycle

One adult, no children -1.78***

2+ adults, no children -0.461***

2+ adults, youngest child 0-5 yrs 0.353***

One adult, youngest child 6-15 yrs 0.795**

2+ adults, youngest child 6-15 yrs -1 91***

One adult, retired, no children -1.6***

Total number of HH adult (18 years or older) members 0.336***

Home Ownership (Home is owned) -0.825*** -0.27**

IMMIGRANT RELATED ATTRIBUTES

Hamidreza Asgari et al. / Transportation Research Procedia 25C (2017) 3083—3099 3093

Respondent is Immigrant/Foreign-born 0.133*** 1 779*** 0.187*

Arrival Time

Migrated 20+ yrs ago 1 09*** 0.334***

Table 6. Determinants of Mode Choice Model Outcome for Shopping Trips (continued)

HOVa Public Transitb NMTc

Migrated 10 to 15 yrs ago 0.387** 1.216***

Migrated 5 to 10 yrs ago 0.317* 0.843** 0.684**

Migrated 0 to 5 yrs ago 0 79*** 1.071** 1.38***

LAND USE ATTRIBUTES

Land Use Index Measurements

Entropy Index 1 42*** 0.571***

Measure of Attractiveness 0.187*** 0.116*

NEIGHBOURHOOD ATTRIBUTES

Immigrant Concentration (Location Quotient)

Less than state -0.479*** -0.167*

Up to twice the state 0.148*

Up to thrice the state -0.181*** 0.964*** 0.178**

Up to four times or more 0.868*** 0.617***

Neighbourhood Typologies

Low social status and high family bonding 0.568***

High social status and high family bonding -0.647* -0.134***

High social status and low family bonding 0.1*

Initial Assessment Statistics

N (Trips Records) 20058

Null log-likelihood -27806

Cte log-likelihood -17234

Final log-likelihood -15287

Rho-square 0.45

Adjusted rho-square 0.448

a,b,c SOV was taken as the base mode * Statistically significant at the 10% level (i.e., p<0.10); ** Statistically significant at the 5% level (i.e., p<0.05); *** Statistically significant at the 1% level (i.e., p<0.01).

Table 7._Determinants of Mode Choice Model Outcome for Social and Recreational Trips.

HOVa Public Transitb NMTc

Intercept (a) -0.438*** -1.86*** 1 13***

PERSONAL ATTRIBUTES

Age 0.005***

Driver Status -0.565***

Education

High school graduate, include GED 0.221***

Race of HH respondent

African American, Black -0.481*** 1 09*** -0.489***

Others 0.415**

Imputed HH Resident's Race and Ethnicity Combined

Hispanic Black 2.8***

Gender-Male -0.277***

HOUSEHOLD ATTRIBUTES

Household Income Per Year

Income Group 1 (<$5,000-$24,999) -0.113*

Household Hispanic status 0.163*

Type of HH

Detached single house -1.07*** -0.206***

Row house or townhouse 0.196***

HH Size 0.251*** 0.099***

HH Vehicle Numbers -1.13*** -0.162***

HH Life Cycle

One adult, no children -1 19***

2+ adults, youngest child 0-5 yrs 0.936*** 0.693***

One adult, youngest child 6-15 yrs 0.356*** -0.575**

One adult, youngest child 16-21 yrs -0.72***

One adult, retired, no children -1.41*** -0.581***

2+ adults, retired, no children 0.224*** 0.118**

Total number of HH adult (18 years or older) members -0.245***

Home Ownership (Home is owned) -0.898*** -0.43***

IMMIGRANT RELATED ATTRIBUTES

Respondent is Immigrant/Foreign-born 0.312*** 0.271***

Arrival Time

Migrated 20+ yrs ago 0.143*

Table 7._Determinants of Mode Choice Model Outcome for Social and Recreational Trips (continued)

HOVa Public Transitb NMTc

Migrated 15 to 20 yrs ago 0.355** 1 98***

Migrated 10 to 15 yrs ago 0.995*** 0.73**

Migrated 5 to 10 yrs ago 0.352*

LAND USE ATTRIBUTES

Land Use Index Measurements

Entropy Index 1.04***

Job-Housing Balance Mix Index -1.528***

Job density (sq. mile) 1.40E-05*

NEIGHBOURHOOD ATTRIBUTES

Immigrant Concentration (Location Quotient)

Less than state -0.115**

Up to twice the state 0.093*

Neighbourhood Typologies

Low social status and high family bonding 0199*** 0.095*

High social status and high family bonding 0.125** -0.982***

High social status and low family bonding -0.19***

Extremely low social status and low family bonding 1.267** -1.01***

Initial Assessment Statistics

N (Trips Records) 11753

Null log-likelihood -16293

Cte log-likelihood -13214

Final log-likelihood -12541

Rho-square 0.23

Adjusted rho-square 0.227

a,b,c SOV was taken as the base mode * Statistically significant at the 10% level (i.e., p<0.10); ** Statistically significant at the 5% level (i.e., p<0.05); *** Statistically significant at the 1% level (i.e., p<0.01).

6.4. Neighbourhood attributes

The major neighbourhood attributes could be summarized as concentration of immigrants and neighbourhood typologies.

6.4.1. Immigrant concentration

The underlying assumption is that immigrants are often employed in co-ethnic businesses located near ethnic neighbourhoods [Blumenberg, 2009]. This explains how concentrations of immigrant populations in the trip maker's household census tract play a key role when choosing a specific mode. For work trips, results indicate that immigrant concentration has a significant impact on transit and NMT usage. People who live in the lowest levels of immigrant concentrations are less likely to use public transit than driving alone. The increase of immigrant concentration in the household census tract increases the likelihood of using both public transit and NMT.

6.4.2. Neighbourhood typologies

Neighbourhood typologies consider both social status and family bonding. High social status individuals have shown a propensity toward driving alone. People with a higher social status possess higher levels of educational qualifications and economic compositions. Furthermore, they possess higher rates of vehicle ownership and prefer driving alone over public transit and HOV usage. In general, people with higher economic and social backgrounds have a tendency to avoid public transit, while lower status people do not. For work trips, the reported coefficients for high family bonding individuals but with different social status are 0.310 and -0.174 for low and high status, respectively, indicating a shift in preference from carpooling to driving alone. Moreover, there are exceptions based on family bonding and trip purposes. For social and recreational trips, high family bonding individuals prefer ride sharing or carpooling over SOV. High family status and low family bonding individuals prefer NMT over SOV for all trip purposes. These types of individuals probably have short commute distances and usually have no dependents.

Low family status individuals prefer alternate modes over driving alone, irrespective of trip purposes. Among alternate modes, they are inclined toward using public transit and carpooling for work trips. Low social status and higher family bonding people prefer HOV, as they like to travel around with their family members or have a lower rate of private vehicle ownership. The change in extremely low status and bonding move their preference towards driving alone for other trips. People who are alone with a low level of family bonding and who live in extremely low status areas may prefer to drive alone for other trips that involve concern for safety or security. Individuals from extremely low social status and bonding prefer public transit and driving alone for recreational trips. In short, people living in areas with a higher social status always prefer to drive alone over HOV or public transit usage, though the preference leans toward the latter two alternative modes with the lowering of status and the increase of family bonding.

7. Conclusion

This paper is an attempt to investigate immigrants' travel behaviour as they constitute a large portion of Florida's population. In particular, Multinomial logit models were developed in order to explore immigrants' mode choice behaviour. The major essence of this paper is two-fold: First, this paper targets neighbourhood typologies as a determinant factor in mode choice decisions. Accordingly, typologies are not merely based on geographical proximity, but also include similarity of socio-demographics. Second, it assumes that the duration of stay in US, which is a sign of cultural assimilation, has significant contribution to immigrants' travel behaviour.

The results of this study is expected to produce a fundamental understanding on how to incorporate neighbourhood effects and behavioural assimilation in understanding the mode choice behaviour of immigrants. The key findings of this paper are listed as follows:

• Irrespective of any trip purposes, larger families and families with more adult members are more likely to take alternative modes. Same pattern is observed for residents in apartment/condominiums and row/town houses. On the contrary, if the trip maker owns a house, lives in a detached single home, has high annual income, or comes from a family with high auto ownership, then he/she is more prone to choose driving alone over alternate modes. Hispanic households show a tendency to carpooling/ ridesharing. Household life cycle

was shown to be a significant indicator in all models, though past research hardly integrated this household attribute.

• Foreign-born individuals are more likely to use alternate modes of transportation than US-born individuals, and their migrating period in the US has significant influence on their mode choice behaviour. Foreign-born immigrants, excluding those that arrived 10 to 15 years ago, always prefer alternative modes for all trip purposes. The most economically stable group prefers driving alone for work trips over alternate modes. However, for shopping and social-recreational trips, they are more likely to use alternate modes over SOV.

• Immigrant quotient was redefined in this study. The new method of grouping provides better insight on travel behaviour. The propensity of transit and NMT usage increases with the increase of immigrant percentage in the household census tract.

• Individuals' mode choice for trips is influenced by the neighbourhood typologies, i.e., social status and family bonding. They prefer driving alone or using alternate modes, depending on high or low status, respectively. High family bonding for high social status favours carpooling for social and recreational trips. Extremely low family status and low family bonding individuals prefer driving alone.

In general, the results of this study present a more challenging way to tackle the diversified complexity of travel behaviour, especially for non-US-born individuals in the US. Overall, the results highlight the importance of new variables in understanding immigrants' mode choice behaviour. For any trip purpose, each choice of travel mode highly depends on compatibility with different land use patterns, auto ownership, and people of versatile backgrounds along with characteristics, social status, and family ties. The estimates of transit ridership and the alternative mode usage can be used in future transportation policy planning relevant to immigrant population at the regional or state level.

References

2009 National Household Travel Survy FDOT File. 2009, U.S. Department of Transportation, Federal Highway Administration.

2009 National Household Travel Survy User Guide. 2009, U.S. Department of Transportation, Federal Highway Administration. http://nhts.ornl.gov/2009/pub/UsersGuideV1.pdf (accessed June 2014).

Alba, R.D., J.R. Logan, and B.J. Stults, The changing neighborhood contexts of the immigrant metropolis. Social Forces, 2000. 79(2): p. 587-621.

Asgari, H., Jin, X. & A. Mohseni, 2014. 'Choice, Frequency, and Engagement - A Framework for Telecommuting Behavior Analysis and Modeling'. Transportation Research Record: Journal of the Transportation Research Board, No. 2413, pp. 101-109.

Asgari, H., and X., Jin, 2015. 'Towards a Comprehensive Telecommuting Analysis Framework; Setting the Conceptual Outline", Transportation Research Record, Journal of Transportation Research Board, No. 2496, pp.1-9.

Bagley, M.N. and P.L. Mokhtarian, The impact of residential neighborhood type on travel behavior: a structural equations modeling approach. The Annals of Regional Science, 2002. 36(2): p. 279-297.

Batalova, J., Immigrants in the U.S. labor force, in U.S. Immigration Trends. 2014, Migration Policy Institute: Washington, DC.

Ben-Akiva M., Lerman S., 1985, Discrete choice analysis, The MIT Press, Cambridge Massachusetts.

Blumenberg, E. and M. Smart, Brother can you spare a ride? Carpooling in immigrant neighbourhoods. Urban Studies, 2014. 51(9): p. 1871-1890.

Blumenberg, E. and M. Smart, Getting by with a little help from my friends... and family: immigrants and carpooling. Transportation, 2010. 37(3): p. 429-446.

Blumenberg, E. and M. Smart, Travel in the 'hood: ethnic neighborhoods and mode choice. University of California Transportation Center, 2009.

Blumenberg, E., Immigrants and transport barriers to employment: The case of Southeast Asian welfare recipients in California. Transport Policy, 2008. 15(1): p. 33-42.

Blumenberg, E., Moving in and moving around: immigrants, travel behavior, and implications for transport policy. Transportation Letters, 2009. 1(2): p. 169-180.

Burbidge, S.K., Foreign living experience as a predictor of domestic travel behavior. Journal of Transport Geography, 2012. 22: p. 199-205.

Casas, J., C. Arce, and C. Frye, Latino immigration and its impact on future travel behavior, in National Household Travel Survey Conference: Understanding Our Nation's Travel, Washington, DC. 2004.

Cascetta E., 2009. "Transportation Systems Analysis: Models and Application". Springer Science + Business Media, LLV, New York.

Crane, R. and R. Crepeau, Does neighborhood design influence travel?: A behavioral analysis of travel diary and GIS data. Transportation Research Part D: Transport and Environment, 1998. 3(4): p. 225-238.

Eisenhauer, E., Angee, A., Hernandez, C. and Y. Zhang, 2007. Immigrants in Florida: characteristics and contributions. 2007, Research Institute for Social and Economic Policy (RISEP), FIU: Miami, FL, USA.

Enam, A. and C.F. Choudhury, Methodological issues in developing mode choice models for Dhaka, Bangladesh. Transportation Research Record, 2011. 2239(1): p. 84-92.

Handy, S., et al., Travel behavior of Mexican and other immigrant groups in California. Berkeley Planning Journal, 2008. 21(1): p. 1-24.

Khalilikhah, M., Habibian, M., and Heaslip, K, Acceptability of increasing petrol price as a TDM pricing policy: A case study in Tehran. Transport Policy, 2016. 45: p. 136-44.

Liu, C.Y. and G. Painter, Travel Behavior among Latino immigrants: the role of ethnic concentration and ethnic employment. Journal of Planning Education and Research, 2012. 32(1): p. 62-80.

Lovejoy, K. and S. Handy, Social networks as a source of private-vehicle transportation: The practice of getting rides and borrowing vehicles among Mexican immigrants in California. Transportation Research Part A: Policy and Practice, 2011. 45(4): p. 248-257.

Martin, P. and E. Midgley, Population bulletin update: immigration in America 2010. 2010, Population Reference Bureau.

Massey, D.S., Arango, J., Hugo, G., Kouaouci, A., Pellegrino, A. and E.J. Taylor. Theories of international migration: a review and appraisal. Population and Development Review, 1993: p. 431-466.

Passel, J.S. and D.V. Cohn, A Portrait of unauthorized immigrants in the United States. 2009, Pew Hispanic Center: Washington, DC.

Saha, D., Development of a mode chice model considering heterogeneous traffice of Dhaka city, in Civil Engineering. 2010, Bangladesh University of Engineering and Technologhy: Bangladesh.

Shams, K., Jin, X., Asgari, H., and Md. Hossan, 2015. 'An Investigation of Temporal Transferability Comparing trip-based and Tour-Based Choice Models'. TRB 94th Annual Meeting Compendium of Papers, Paper#15-0287.

Sharifi, M.S., Stuart, D., Christensen, K.M., Chen, A., Kim, Y.S., Chen, Y. (2014). Analysis of walking speeds involving individuals with disabilities in different walking environments. Proceeding of the Transportation Research Board 93rd Annual Meeting, Washington DC., January 2014.

Sharifi, M.S., Stuart, D., Christensen, K.M., Chen, A. (2015). Exploring traffic flow characteristics and walking speeds of heterogeneous pedestrian stream involving individuals with disabilities in different walking environment. Proceeding of the Transportation Research Board 94th Annual Meeting, Washington DC., January 2015.

Sharifi, M.S., Stuart, D., Christensen, K.M., Chen, A. (2016). Time headway modeling and capacity analysis of pedestrian facilities involving individuals with disabilities. Proceeding of the Transportation Research Board 95th Annual Meeting, Washington DC., January 2016.

Smart Location Data Base. U.S. EPA. http://www.epa.gov/smartgrowth/smartlocationdatabase.htm (accessed on June 2014).

Soltani Sobh, A., Heaslip, K., Barnes, R., Bosworth, R. Yook, D., 2016c. An Aggregated Panel Data Analysis to Model Electric Vehicle Adoption Rates. In Transportation Research Board 95th Annual Meeting, Washington D.C., paper # 16-4010.

Sobh, Ali Soltani, 2015. Transportation Economics and Energy. All Graduate Theses and Dissertations. Paper 4165. http://digitalcommons.usu.edu/etd/4165.

Smart, M.J., A nationwide look at the immigrant neighborhood effect on travel mode choice. Transportation, 2014. 41: p. 1-21.

Smart, M.J., US immigrants and bicycling: Two-wheeled in Autopia. Transport Policy, 2010. 17(3): p. 153-159.

Tal, G. and S. Handy, Travel behavior of immigrants: An analysis of the 2001 National Household Transportation Survey. Transport Policy, 2010. 17(2): p. 85-93.

Train, K., Discrete Choice Methods with Simulation, Cambridge University Press, 2003.

U.S. Census Bureau. 2010 American Community Survey. http://factfinder2.census.gov/ (accessed June 2014).

Vaziri, M., Dashtestaninejad, H., Asgari, H., and F., Salimi, 2014. 'Rural Travel Behavior in Developing Countries: A Study of Farm-Based Villages in Iran'. The Journal of Sustainable Mobility, Vol.1, No.2, pp. 25-45 (21).

Washington, S.P., Karlaftis, M.G. and F.L. Mannering, 2011. "Statistical and Econometric Methods for Transportation Data Analysis", Second Edition. Taylor & Francis Group, LLC, Boca Raton.