3.

3 Quantitative Research

In this study, factors influencing the intention to use ride - hailing services were
examined, specifically targeting those who have previously used such services. The
research data was collected through an online survey on the intention of Ho Chi Minh
City students to use ride - hailing services. To encourage participation, small gifts were
provided as a token of appreciation. Data collection spanned nearly three months, from
July 10 to September 31, 2024.

Participants were invited to join the survey and agreed to do so, driven by various
motivations, such as a desire to deepen their knowledge in this field, an interest in the
research topic, or a sense of responsibility towards the research team. Additionally,
participants might recognize personal benefits from their involvement, such as gaining
relevant study materials for their courses or acquiring a deeper understanding of the topic.
Moreover, some may find satisfaction in contributing to knowledge that could enhance
products, services, or policies. We also planned to encourage responses with rewards,
such as coffee shop vouchers or study materials relevant to general education subjects, to
stimulate interest.

3.3.1 Research Sample Design

Sampling Method: This study utilized a convenience sampling method, allowing the
research team to select easily accessible participants.

Sample Population: University students who have used ride - hailing services in Ho Chi
Minh City.

Sample Size: Based on Cohen's recommendation of 80% statistical sensitivity and 5%

significance level, the proposed research model requires at least 137 survey samples to
achieve a minimum R-squared value of 0.10 (Hair, 2014). However, to minimize the risk
of incomplete or unclear responses, the research team increased the sample size to 200
participants, improving the reliability and accuracy of the results, while minimizing the
impact of missing or ambiguous data on the model.

3.3.2 Questionnaire Design

Section 1: Screening Question

Question 1: Have you ever used ride - hailing services?

○ Yes (proceed with the survey)

○ No (You may exit the survey. Thank you for your participation)

Section 2: Survey Content

The survey consists of 28 questions rated on a 5-point Likert scale to assess factors such
as perceived service quality, attitude, subjective norms, perceived behavioral control,
perceived risk, perceived usefulness, perceived ease of use and usage intention among
those who have used ride - hailing services. The Likert scale is used to evaluate each item,
where 1 represents "strongly disagree" and 5 represents "strongly agree."

1: Strongly disagree

2: Disagree

3: Neutral

4: Agree

5: Strongly agree

Section 3: Demographic Information

Personal information collected includes gender, age, occupation, and income.

3.3.3 Data Analysis Method

Upon completing data collection, the data will be organized, classified, and any invalid
samples will be removed. The research team will then use SmartPLS 4 software to
perform analysis steps, including data description, reliability testing (Cronbach’s Alpha),
convergent validity testing, discriminant validity testing of constructs within the model,
structural model accuracy evaluation (VIF), and hypothesis testing.
[Link] Observed Variable Quality Assessment
In A Primer on Partial Least Squares Structural Equation Modeling (2014), Hair and
colleagues suggest that an outer loading value of 0.7 or above indicates a strong indicator.
Outer loading values below 0.7 are considered weak and may not reliably measure latent
variables in the PLS-SEM model. Indicators with outer loading values below 0.7 may
lack sufficient strength to accurately measure the corresponding latent variable.
Evaluating the quality of observed variables involves assessing whether any indicators
within a factor structure poorly represent the latent parent variable. Weak indicators are
removed to retain only the most representative indicators for the latent variable.
The research team used the PLS-SEM algorithm to determine outer loading coefficients.
Hair and colleagues recommend that observed variables are valid when their outer loading
is 0.7 or higher. Indicators with an outer loading below 0.4 should be removed from the
model. For outer loadings between 0.4 and 0.7, the decision to remove or retain depends
on other indicators, such as composite reliability (CR) and the corresponding factor's
convergent validity (e.g., AVE score).
[Link] Reliability, Convergent Validity, and Discriminant Validity of Constructs in
the Model
Reliability:
SMARTPLS 4 provides three indicators to evaluate the reliability of scales or internal
consistency: Cronbach's alpha, Composite reliability rho_a, and Composite reliability
rho_c (CR). Hair et al. (2017) recommend focusing on two primary indicators: Cronbach's
alpha and Composite reliability rho_c.
According to these authors, Cronbach's alpha, a traditional reliability measure, has certain
limitations and often underestimates the true reliability of the scale. Instead, using the
composite reliability coefficient rho_c is considered a more appropriate method for
evaluating reliability.
The approach to evaluating composite reliability is similar to Cronbach's alpha. This
coefficient ranges from 0 to 1 (values below 0 indicate data errors); the closer it is to 1,
the higher the reliability. Composite reliability between 0.6 and 0.7 is acceptable for
exploratory studies, while the optimal range is between 0.7 and 0.9 (Nunally & Bernstein,
1994). If composite reliability exceeds 0.95, it may indicate redundancy among observed
variables.
Cronbach's alpha generally underestimates reliability, while composite reliability may
overestimate it.
Composite reliability values between 0.60 and 0.70 are acceptable for exploratory
research, whereas values between 0.70 and 0.90 are adequate for more advanced research
stages.
Cronbach's alpha ≥ 0.7 (DeVellis, 2012) and Composite reliability ≥ 0.7 (Hair et al.,
2013) are considered satisfactory.
Convergent Validity:
Convergent validity is assessed using the Average Variance Extracted (AVE). An AVE of
0.5 or higher is considered indicative of convergent validity (Fornell & Larcker, 1981),
meaning that on average, the latent variable explains at least 50% of the variance in its
observed variables (Hock & Ringle, 2010).
If a scale has an AVE below 0.5, the research team will iteratively remove observed
variables with the lowest outer loadings to raise the AVE value. This process is repeated
until the factor achieves convergent validity. If, after this process, the factor retains at
least two observed variables and achieves AVE ≥ 0.5, the scale is deemed to have
convergent validity. If not, the factor violates convergent validity and should be removed
from the model.
Discriminant Validity:
Discriminant validity of the constructs is established when the square root of the average
variance extracted for a construct is greater than its correlation with other constructs in the
model. Previous studies commonly use the Fornell-Larcker criterion to assess
discriminant validity. Therefore, this study utilizes the Fornell-Larcker index to ensure the
distinctiveness of each construct.
[Link] Model Structure Evaluation
Multicollinearity or collinearity issues in the model can lead to biased regression
coefficients and significant p-values, resulting in inaccurate conclusions about
relationships within the model. Hair et al. (2019), When to Use and How to Report the
Results of PLS-SEM. European Business Review, recommend VIF (variance inflation
factor) thresholds for assessing collinearity:
VIF ≥ 5: high likelihood of collinearity; model severely affected.
3 ≤ VIF < 5: possible collinearity in the model.
VIF < 3: no collinearity issue in the model.
PLS-SEM aims to maximize the R2 values of endogenous latent variables in the path
model. Higher R2 values are preferable, although specific values depend on the model
context and research field. Generally, values of 0.75, 0.50, and 0.25 indicate substantial,
moderate, and weak explanatory power, respectively.
Effect Size (f²): Cohen (1988) suggests the following thresholds for assessing the impact
of independent variables:
f² < 0.02: negligible or no effect.
0.02 ≤ f² < 0.15: small effect.
0.15 ≤ f² < 0.35: moderate effect.
f² ≥ 0.35: large effect.
Predictive Accuracy (Q²): Hair et al. (2019) provide Q² thresholds for model predictive
accuracy as follows:
0 < Q² < 0.25: low predictive accuracy
0.25 ≤ Q² ≤ 0.5: moderate predictive accuracy
Q² > 0.5: high predictive accuracy
 CHAPTER 4: DATA ANALYSIS AND RESULTS

4.1 Sample and data collection

A total of 219 online surveys were distributed to users who have used or are using ride-
hailing services through a Google form.
The survey results showed that among the participants, there were 166 females (75.8%)
and 53 males (24.2%), indicating a higher proportion of females than males. Although
there was a gender gap in the survey sample, this was considered acceptable because
gender did not significantly affect the intention to use ride-hailing services. The majority
of respondents were second-year students (over 56%) because the author, who is also a
second-year student, took advantage of relationships to collect information. At the same
time, the survey also collected random responses from students of other universities, in
which the proportion of fourth-year students only accounted for no more than 4%.
Notably, up to 84.5% of students regularly use technology car services, while only 34%
do not use them regularly, showing that students have a very high demand for this service
because of its convenience and usefulness
frequency Percentage Accumulated
Percentage
Sex
Female 166 75.8 75.8
Male 53 24.2 100.0
Total 219
Category
First-year 14 6.4 6.4
Second-year 124 56.6 63,0
Third-year 41 18.7 81,7
Fourth-year 30 13.7 95,4
Above fourth-year 10 4.6 100.0
Total 219

Usage Frequency
Frequent 185 84.5 84.5
Infrequent 34 15.5 100.0
Total 219

Table 9: Survey sample statistics

4.2 Assessment of the reliability of the scale (Cronbach’s Alpha)
Based on the presented theory, the research team conducted an examination of the outer
loading index of 29 observed variables. The results showed that all observed variables had
an outer loading coefficient greater than 0.7, ensuring the significance of the observed
variables and the reliability of the scales met the requirements of internal consistency.

Observed Variable Outer

Loading
Perceived 1 I believe that the ride-hailing app 0.799
Service provides quick responses to
Quality customers
2 I believe that the ride-hailing app 0.86
has a reliable technological platform
3 The ride-hailing app is safe as it 0.889
provides full information about
drivers to customers
4 The ride-hailing app offers 0.853
economic benefits as it provides
lower costs for customers
Attitude 1 I like the ride-hailing app 0.89
2 I find the ride-hailing app useful 0.833

3 I find the ride-hailing app enjoyable 0.919

4 I feel happy when using the ride- 0.863

hailing app
Subjective 1 Most people important to me support 0.871
Norm my use of the ride-hailing app
2 Most people important to me agree 0.908
that I use the ride-hailing app
 3 Most people important to me 0.882
encourage me to use the ride-hailing
app
Perceived 1 I am capable of using the ride- .75
Behaviora hailing app
l Control
2 If I want, I can easily use the ride- 0.929
hailing app
3 I have sufficient time and energy to 0.835
use the ride-hailing app
Perceived 1 I feel unsafe providing my 0.891
Risk authentication information to the
ride-hailing app
2 I feel my privacy would be 0.868
compromised by using the ride-
hailing app
3 I feel there would be a data security 0.904
risk when using the ride-hailing app
Perceived 1 Using the ride-hailing app helps me 0.847
Usefulness save time
2 Using the ride-hailing app improves 0.769
my quality of life
3 Using the ride-hailing app ensures 0.75
safety
4 Using the ride-hailing app provides 0.771
convenience
5 Using the ride-hailing app helps me 0.769
save costs
Perceived 1 I believe the functions of the ride- 0.822
 Ease of hailing app are clear
Use 2 I believe I can easily learn to use the 0.792
ride-hailing app
3 I believe I can quickly become 0.841
proficient in using the ride-hailing
app
4 The process of conducting 0.786
transactions with the ride-hailing
app is easy
Usage 1 I feel it is necessary to use the ride- 0.887
Intention hailing app in my daily activities
2 I plan to use the ride-hailing app in 0.891
the near future
3 I intend to use the ride-hailing app in 0.882
the future
Table 10: Quality of observed variables

4.3 Assessment of reliability, validity, and discriminant validity of the constructs in

the model.

From Table 11, it can be seen that all factor structures have good reliability, determined
by two indicators: Cronbach's alpha reliability coefficient and composite reliability
coefficient (rho_c), both of which have values greater than 0.7. In addition, the average
variance extracted (AVE) of the scales all exceed 0.5, which proves that the research
structural scales have achieved high convergent validity.
Scale Composite Reliability (CR) Average Variance
Extracted (AVE)
Perceived Service Quality 0.872 0.724
 Attitude 0.899 0.769

Subjective Norm 0.865 0.787

Perceived Behavioral Control 0.831 0.708

Perceived Risk 0.865 0.788

Perceived Usefulness 0.841 0.612

Perceived Ease of Use 0.828 0.657

Usage Intention 0.864 0.786

Table 11: Reliability rating table, AVE

To assess discrimination, the research team used the popular method in previous
research, Fornell - Larcker. The results from the analysis showed that all factors in the
model ensured discrimination, when there was no correlation coefficient between any pair
of factors exceeding the square root value of the average variance extracted (AVE) of
those factors.

CCQ CLDV DSD HI ID KSHV RR TĐ

CCQ 0.887
CLDV 0.362 0.851
DSD 0.403 0.553 0.811
HI 0.555 0.551 0.561 0.782
ID 0.570 0.734 0.663 0.703 0.887
KSHV -0.155 -0.132 -0.182 -0.145 -0.173 0.841
RR -0.413 -0.636 -0.459 -0.585 -0.646 0.040 0.888
TĐ 0.502 0.572 0.568 0.519 0.675 -0.202 -0.402 0.877

Table 12: Discriminant value table

4.4 Model evaluation

In Table 4.4, the VIF coefficients of the variables DSD, TĐ, CCQ, KSHV, RR, HI for the
dependent variable ID are all below 3, which shows that the model does not have
multicollinearity.
The coefficient of determination R2 of the dependent variable ID reached 0.755, proving
that exogenous variables explain endogenous variables well.
Considering the results from the f2 index in Table ..., the variable CLDV has a medium
impact, while the variable KSHV has a very small impact or almost no impact. The
remaining variables have a small impact. This shows that the level of explanation of the
independent variables for the dependent variable is quite low.
Finally, for the dependent variable "Intention to use", the value Q2=0.712 shows that the
predictive level of the independent variables for the dependent variable is high.

Scale VIF R2 f2 Q2
Perceived Service Quality 2.247 0.150

Attitude 1.962 0.073

Subjective Norm 1.625 0.047

Perceived Behavioral Control 1.064 0.001

Perceived Risk 2.001 0.041

Perceived Usefulness 2.158 0.315 0.070 0.304

Perceived Ease of Use 1.838 0.062

Usage Intention 0.755 0.712

Table 13: Two-variable relationship

4.5 Testing research hypotheses

Relations Hypothe Origin Mean t-value p-

hip sis al Weigh value
Weight t

(O) (M)

PEU -> H1 0.167 0.1 3.47 0.0 Support

UI 66 2 01 ed

AT -> UI H2 0.188 0.1 3.02 0.0 Support

88 5 03 ed

PSQ -> H3 0.287 0.2 4.16 0.0 Support

UI 86 8 00 ed
 PBC -> H4 -0.012 -0.017 0.34 0.7 Rejecte
UI 5 30 d

PR -> UI H5 -0.142 -0.141 2.69 0.0 Support

6 07 ed

SN -> UI H6 0.137 0.1 2.74 0.0 Support

41 4 06 ed

PU -> UI H7 0.192 0.1 2.48 0.0 Support

90 5 13 ed

PEU -> H8 0.561 0.5 11.2 0.0 Support

PU 65 74 00 ed

Table 14: Testing hypothesis

The results from Table 4 show that the original weights are significant for the
bootstrapping mean weights, as all the weights are within the 95% confidence interval.
This shows that the estimates in the model are reliable. At the same time, there are 6
factors that have a positive impact on customer satisfaction at the 5% and 1% statistical
significance levels. Specifically, hypotheses H1, H2, H3, H5, H6, H7 and H8 are
accepted, while hypothesis H4 has a p value of 0.730 (greater than 0.05), indicating that
this relationship is not statistically significant. Among the influencing factors, Perceived
Service Quality has the strongest impact on customers' intention to use ride-hailing (β =
0.287), followed by Perceived Usefulness (β = 0.192), Attitude (β = 0.188), Perceived
Ease of Use (β = 0.166), and Subjective Norm (β = 0.177). In particular, the β coefficients
of Perceived Behavioral Control and Perceived Risk have negative values, β = -0.012 and
β = -0.142, respectively.
It is worth noting that the estimation results do not find statistical evidence for the
relationship between Perceived Behavioral Control and Intention to Use Ride-hailing
Application, so hypothesis H4 is rejected.
4.6 Conclusion
The research results show that Perceived Service Quality (PSQ) is the factor that has the
strongest impact on the intention to use ride-hailing services, with a coefficient of β =
0.287. This indicates that users value the reliability, speed and safety that ride-hailing
applications bring, thereby directly affecting their decision to choose the service. These
results are similar to previous studies on service quality in the field of online services,
emphasizing the importance of this factor in building trust and promoting consumers'
service usage behavior.
The research results show that many factors significantly affect users' intention to use
ride-hailing services. Specifically, Perceived Usefulness (HI) is the factor with the most
obvious impact, with a coefficient of β = 0.192. This shows that users perceive practical
benefits from the application, such as saving time, costs and improving the quality of life.
This is an important factor that application developers need to focus on, in order to
optimize the features and benefits that users can receive from the service.
User attitude (TD) towards the application, with a coefficient of β = 0.188, also has a
positive influence on the intention to use. This attitude is often formed from positive
experiences with the product, so improving the user experience is necessary to build a
positive attitude and thereby promote the decision to use the service.
Perceived ease of use (DSD) with a coefficient of β = 0.167 is the next factor with an
important impact. Users tend to use the service more if the application is easy to use and
has a friendly interface. Therefore, user interface (UI) design and user experience (UX)
optimization are key factors in increasing service usage rates.
Subjective norm (CCQ), with β = 0.137, reflects the influence of important people around
users on the decision to use the service. This shows that social factors, including the
influence of friends, family or community, play an important role in encouraging users to
use the service.
In particular, Perceived behavioral control (KSHV) and Perceived risk (RR) have
negative effects on usage intention, with β coefficients of -0.012 and -0.142, respectively.
This indicates that Perceived risk, such as concerns about security and privacy, has a
significant negative impact on the decision to use the service. Users will be less inclined
to use the service if they feel insecure about their personal data. However, Perceived
behavioral control did not have a significant impact on the intention to use the service.
This may be explained by the development of today's ride-hailing platforms, where the
user experience has been optimized, reducing the need for self-control or user intervention
during use.
The results of the study did not support the hypothesis of the influence of Perceived
behavioral control, suggesting that the ability of users to self-control and regulate their
behavior when using the service is not a decisive factor in the context of ride-hailing
services. This may be the result of these applications simplifying the usage process,
making users less likely to control their behavior.
In summary, factors such as service quality, usefulness, attitude, and perceived ease of use
all have a strong influence on the intention to use ride-hailing services. Meanwhile,
perceived risk needs to be controlled to avoid reducing users' usage. These results provide
valuable information for ride-hailing service developers to improve their products and
marketing strategies to attract and retain users more effectively.
 APPENDIX
Outer loadings
KSH
CCQ CLDV DSD HI ID V RR TĐ
CCQ1 0,871
CCQ2 0,908
CCQ3 0,882
CLDV1 0,799
CLDV2 0,860
CLDV3 0,889
CLDV4 0,853
DSD1 0,822
DSD2 0,792
DSD3 0,841
DSD4 0,786
HI1 0,847
HI2 0,769
HI3 0,750
HI4 0,771
HI5 0,769
ID1 0,887
ID2 0,891
ID3 0,882
KSHV1 0,750
KSHV2 0,929
KSHV3 0,835
RR1 0,891
RR2 0,868
RR3 0,904
TĐ1 0,890
TĐ2 0,833
TĐ3 0,919
TĐ4 0,863

R-square

R-square R-square adjusted

HI 0,315 0,312
ID 0,755 0,747
f-square

f-square
CCQ -> ID 0,047
CLDV -> ID 0,150
DSD -> HI 0,460
DSD -> ID 0,062
HI -> ID 0,070
KSHV -> ID 0,001
RR -> ID 0,041
TĐ -> ID 0,073

Fornell-Larcker criterion
KSH
CCQ CLDV DSD HI ID V RR TĐ
CCQ 0,887
CLDV 0,362 0,851
DSD 0,403 0,553 0,811
HI 0,555 0,551 0,561 0,782
ID 0,570 0,734 0,663 0,703 0,887
-
KSHV -0,155 -0,132 -0,182 -0,145 0,173 0,841
-
RR -0,413 -0,636 -0,459 -0,585 0,646 0,040 0,888
-
TĐ 0,502 0,572 0,568 0,519 0,675 0,202 -0,402 0,877

Construct reliability and validity

Average
Composite Composite variance
reliability reliability extracted
Cronbach's alpha (rho_a) (rho_c) (AVE)
CCQ 0,865 0,876 0,917 0,787
CLDV 0,872 0,873 0,913 0,724
DSD 0,828 0,836 0,884 0,657
HI 0,841 0,844 0,887 0,612
ID 0,864 0,864 0,917 0,786
KSHV 0,831 1,054 0,878 0,708
RR 0,865 0,866 0,918 0,788
TĐ 0,899 0,904 0,930 0,769

Inner model
CLD
CCQ V DSD HI ID KSHV RR TĐ
CCQ 1,000
CLDV 1,000
DSD 1,000 1,000
HI 1,000
ID
KSHV 1,000
RR 1,000
TĐ 1,000

Standard T statistics
Original Sample deviation (|O/STDEV
sample (O) mean (M) (STDEV) |) P values
CCQ -> ID 0,137 0,141 0,050 2,744 0,006
CLDV -> ID 0,287 0,286 0,069 4,168 0,000
DSD -> HI 0,561 0,565 0,050 11,274 0,000
DSD -> ID 0,167 0,166 0,048 3,472 0,001
HI -> ID 0,192 0,190 0,077 2,485 0,013
KSHV -> ID -0,012 -0,017 0,034 0,345 0,730
RR -> ID -0,142 -0,141 0,053 2,696 0,007
TĐ -> ID 0,188 0,188 0,062 3,025 0,003