Page 44 - IJPS-9-1

International Journal of
Population Studies Transportation assimilation in Hong Kong

5 is our final model, which dissects the total effects into Equation V is our final model, in which we streamlined
initial effects and long-term effects of our key independent previous models by keeping only those more important
variables. items for i’s likelihood to take a minibus. Throughout
Equations I – V below correspond to Models 1 – 5 in the analysis, we applied linear regressions and robust
the Results section. We started with Equation I, in which standard errors clustered in residential districts. Below we
we only included immigrant, i’s arrival, years of duration in further explain our reasons for including arrival cohorts
Hong Kong, ethnic background, Cantonese ability, i’s index and a series of locational- and temporal-fixed effects and
of interaction (represented by P*), and a series of control applying linear approximation in our analysis.
variables. Minibus is the dependent variable capturing Minibus MigAge Duration Ethnicity
i
i
i
i
immigrant i’s probability of choosing minibus over other i Cantonese P Income Duration
*
modes of transportation in journeys to work, and ε is the i i i i
i
individual-level robust standard error. MigAge Duration Ethnicity
i
i
i
Minibus MigAge Duration Ethicity MigAge Ethnicityy Duraiton
i
i
i
i
i
i
i
*
Cantonese P Controlss i (I) MigAge Ethnicity Duration
i
i
i
i
i
Income Controlss i (V)
i
i
In Equation II, we further included two interactions,
which are individual i’s years of staying in Hong Kong and Following classical immigration research on
age at arrival and individual i’s years of staying in Hong immigrants’ integration, we controlled for one’s arrival
Kong and ethnic group. In Equation III, we included more cohort in our analysis (Martinović, 2013). The major
interactions (including one three-way interaction) to depict advantage of controlling for arrival cohorts is that we
the potential interethnic differences in minibus ridership, can trace the groups of immigrants in cross-sectional
which are individual i’s age at arrival and ethnicity as well as data, in our case, the immigrants arriving in Hong Kong
individual i’s years of staying in Hong Kong, age at arrival, in the same cohort, similarly to tracing individuals in
and ethnicity. In Equation IV, we again expanded our longitudinal data. In this way, we can control for potential
interactions terms by including some important control cohort effects.
variables into the interactions, including the interaction We have also included a series of locational- and
between individual i’s years of staying in Hong Kong and temporal-fixed effects, including residential district-fixed
logged personal income, individual i’s years of staying in effects, working district-fixed effects, residential-working-
Hong Kong and Cantonese ability, and individual i’s years district fixed effects, census year-fixed effects, year-
of staying in Hong Kong and gender. residential-district fixed effects, and year-working-district
Minibus MigAge Duration Ethnicity fixed effects. By incorporating a series of residential district
i
i
i
i
*
Cantonese P Duratioon MigAge i dummies, we are only comparing the transportation
behaviors of immigrants living in the same district, which
i
i
Duration Ethnicity Controls (II) therefore wipes out the possibility of not taking minibuses
i
i
i
i
as a result of having few minibus routes in certain residential
Minibus MigAge Duration Ethnicity districts. Similarly, by including working district dummies,
i
i
i
i
Cantonese P Duratioon MigAge i we are only comparing the passengers working in the same
*
i
i
Duration Ethnicity MigAge district, which accounts for the possibility of not taking
i
i
i
Ethnicity Duratiion MigAge minibuses to work as a result of not having minibus routes
i
i
i
Ethnicity Controls i (III) in certain working districts. To also control the distance
between one’s residential location and working location,
i
i
Minibus MigAge Duration Ethnicity we included the interactions between residential location
i
i
i
i
dummies and working location dummies. To account for
Cantonese P Income Gender the effects of potential district development throughout
*
i
i i
i
Duration MigAge Duration the years (e.g., any development of the subway system
i
i
i
Ethnicity MigAge EEthnicity throughout the years that may affect minibus ridership),
i
i
i
Duration MigAge Ethnicity we have controlled for census year dummies. In the end,
i
i
i
Duration Cantonesse Duration by incorporating the year-fixed effects and the district-
year-fixed effects (i.e., the interaction between year
i
i
i
EthDen Duration Income dummies and district dummies), we can control potential
i
i
i
Duration Gender Controls i i (IV) time-specific regional factors (e.g., any suspension of the
i
i
Volume 9 Issue 1 (2023) 38 https://doi.org/10.36922/ijps.0386