INTRODUCTION
Overview of Vietnam’s retail sector
Vietnam is becoming an appealing retail market, driven by a youthful demographic, with 70% of the population aged between 15 and 64 The market is anticipated to grow significantly due to rising population and purchasing power Notably, the urban population, which holds the greatest potential for retail, is projected to grow at an annual rate of 2.6% from 2015 to 2020, marking the highest growth in the region (World Bank, 2015; United Nations).
Vietnam's urban population is rapidly growing due to urbanization and rural-to-urban migration, positioning the country as a dynamic emerging economy with improving living standards and rising disposable incomes According to the Boston Consulting Group (2013), Vietnam is experiencing the fastest growth of wealthy and middle-class consumers in the region, with an expected increase from 12 million to 21 million between 2012 and 2020 This demographic, earning over VND 15 million (USD 714) monthly, represents a significant opportunity for retailers targeting potential customers.
Vietnam's improving infrastructure is attracting international retailers, with major players like Metro Cash & Carry, BigC, Emart, Aeon, Takashimaya, and AuchanSuper entering the market This influx, alongside domestic competitors such as Co-op Mart and Vincom Mall, has intensified competition in Vietnam's retail sector According to CBRE (2014), Hanoi and Ho Chi Minh City are among the top 10 Asian cities for retail market expansion, with Hanoi ranked third after Shanghai and Beijing The retail industry in Vietnam has experienced significant growth, with revenues increasing by 60% from 2009 to 2013 and projected to reach US$109 billion by 2017, highlighting its vast potential for new investors.
The two primary retail channels are traditional retail and electronic retailing (e-tailing) Traditional retail involves direct exchanges of goods or services with customers through various outlets, including general stores, convenience stores, supermarkets, and hypermarkets.
The rise of e-tailing, driven by advancements in information technology and internet accessibility, allows customers to shop online by visiting store websites to purchase goods or services Through online shopping, consumers can view product images, explore specifications, and compare prices, all with just a few mouse clicks This convenient method saves time and eliminates the need for physical transportation of products Although e-tailing sales in Vietnam were low from 2013 to 2015, they have been growing rapidly at an annual rate of over 35 percent.
Online retail sales (billion USD) Total retail sales (billion USD)
Figure 1.1: Vietnam’s retail sales and online retail sales from 2013 to 2015
Source: General Statistics Office of Vietnam (2013, 2014, 2015).
Trend in shopping behavior in Viet Nam
The rapid expansion of Information and Communication Technologies (ICT) has significantly transformed consumer behavior and preferences, with individuals increasingly favoring products and services that enhance their convenience Online shopping has emerged as a solution to meet these demands for convenience According to Nielsen (2014), Vietnam ranks third in the ASEAN region, following Singapore and the Philippines, with 58% of the population utilizing mobile phones for online shopping.
Between 2013 and 2015, the internet and social networks like Facebook, Twitter, Zalo, and YouTube significantly transformed community connections in Vietnam Additionally, online shopping platforms such as Lazada, Adayroi, Shoppee, and Tiki have become essential in facilitating informal online shopping, with a substantial portion of Vietnam's e-commerce occurring through these social networks and applications.
The rapid growth of online shopping in Vietnam can be attributed to busy lifestyles, the convenience of e-commerce, and advancements in ICT According to VECITA (2014), online shopping sales revenue in Vietnam reached $2.97 billion, with an average online purchase per capita of $145, making up 2.12% of total retail sales The most popular categories included fashion and cosmetics (28% of online sales), followed by furniture and electronics (25%), and food and beverages (16%) By 2015, VECITA reported a rise in average online spending to $160 per person, with total revenue increasing to $4.07 billion—a 37% growth from 2014—representing approximately 2.8% of total retail sales The leading purchases were clothing, footwear, and cosmetics (32%), alongside technology, household appliances, books, and gifts.
Figure 1.2: Percentage online shoppers of the online accessing in Vietnam
Online retail sales in Vietnam account for 2 to 5% of total sales, which is similar to other regional countries but significantly lower than China (13.5%) and Korea (11.2%) This indicates substantial growth potential for the e-tailing sector in Vietnam Understanding shopper behavior and the factors that motivate online shopping can be crucial for the sector's development.
Research questions and objectives
This study aims to analyze consumer preferences between online and in-store shopping by identifying key attributes influencing their choices It seeks to understand the factors that encourage consumers to shop in physical stores versus those that drive them to opt for online shopping.
This study analyzes the time and monetary costs influencing consumers' decisions between online and in-store shopping channels through a random utility framework Key factors such as price, ordering time, travel time and costs, and delivery time are evaluated, as they significantly impact overall utility According to random utility theory, consumers aim to maximize their utility when selecting between online and physical stores A random utility model incorporating these time and money attributes is utilized to assess the decision-making process for purchasing books.
Consumer choices between online and in-store shopping are influenced by various socio-demographic factors and personal attitudes This study aims to explore four key objectives related to these influences on shopping behavior.
(1) Evaluating the impacts of shopping channel’s attributes on the choice between online shopping versus in-store shopping.
(2) Determining the effects of individual characteristics on the consumer’s choice.
(3) Examining the trade-off between the alternative-specific attributes in the competition between online shopping and in-store shopping, to predict the change in the choice probability.
(4) Measuring the willingless to pay between time and cost attributes of alternatives Therefore, this study raises the following questions:
(1) Which attributes affect on individual’s choice between online shopping and in-store shopping?
(2) How do the social-demographic affect the shopping mode choice?
(3) How does the probability of choosing shopping mode changes when changing the alternative-specific attributes?
(4) How much the consumers are willing to tradeoff between the time and cost attributes?
This thesis employs random utility models to analyze consumer decision-making between online and traditional shopping channels, revealing the key factors influencing their choices The findings will assist sellers in both channels by identifying the critical determinants affecting consumer behavior, which is essential for developing effective retail strategies and predicting valuation indicators.
The thesis provides online sellers with insights into the benefits and challenges they face, including pricing fluctuations and the impact of travel time, costs, and delivery efficiency By understanding these factors, sellers can refine their sales and marketing strategies and identify key obstacles to attracting more customers Additionally, store sellers can recognize significant competitive advantages based on consumer preferences, enhancing their overall market position.
The scope of study
This thesis presents a practical study on consumer preferences between online and in-store shopping, utilizing a random utility model Conducted in August 2017, the research employed both paper-and-pencil surveys and face-to-face interviews to gather data It integrated Revealed Preference and Stated Preference methodologies, focusing on various locations in Ho Chi Minh City, including Book Street Nguyen Van Binh, Walking Street Nguyen Hue, and several universities and cafes The study surveyed 352 respondents, who were decision-makers in book purchases over the past year, while acknowledging constraints in finance and time.
The reason why this research attempts to focus on a single class of product, namely books, because of the three primary reasons:
To minimize the influence of product characteristics, such as sensory or non-sensory attributes, and experience goods like clothing, it's essential to highlight the effects of the shopping channel Additionally, purchasing a specific book from a physical bookstore appears similar to buying it from an online retailer, indicating that the shopping experience may not significantly differ across channels.
2009) Therefore, the effects of product characteristics can be eliminated from the model.
According to the Vietnam E-Commerce Report (2016), books rank fifth among products commonly purchased online, following fashion, IT/mobile items, kitchen/home appliances, and food/drinks This data suggests that many participants in the survey are familiar with online bookstores.
This study included an experiment examining the portfolio of best-selling products frequently purchased online The findings revealed that books rank as the third most popular product category among consumers shopping through online channels.
Organization of the thesis
This thesis is organized into five chapters: Chapter 1 offers an overview of Vietnam's retail sector, exploring shopping behavior trends, research questions, objectives, and the study's scope Chapter 2 reviews existing literature, categorizing it into four analytical approaches: binary choice model, online shopping frequency, random utility model, and structural equation model Chapter 3 outlines the research methodology, detailing the choice model and data collection methods through revealed and stated preferences, including experimental design, sample size, and estimation techniques Chapter 4 provides a statistical description and presents empirical results, while Chapter 5 concludes with policy implications, acknowledges limitations, and suggests directions for further research.
LITERATURE REVIEW
Binary Choice models
Binary choice models are used to analyze the choice between two alternatives (Train, 2002).
When analyzing online shopping decisions, the two primary alternatives are online and in-store shopping The outcome variable is binary, taking the value of "one" when a decision maker engages in the desired action, and "zero" when the action is not taken In these models, the explanatory variables focus on the attributes of the decision makers, excluding any attributes of the shopping alternatives themselves.
This research examines two shopping modes: online and in-store shopping When a decision-maker opts for online shopping, it is assigned a value of one (� = 1), indicating a preference for this mode Conversely, if the individual chooses in-store shopping, the value is zero (� = 0), reflecting the absence of an online shopping choice The model subsequently defines the probability of purchasing an item online based on these values.
Pr(� = 1) = �(�|�) where � is the parameter to be estimated, and � is the vector of explanatory variables.
In a binary choice scenario, the probability of a decision maker or individual consumer choosing between online and in-store shopping can be succinctly represented, as outlined by Ben-Akiva and Lerman (1985).
� � denotes the indirect utility of shopping in-store, while � � is the utility from shopping online Dividing both nominator and denominator by � � � , we have
If one assume � � − � � = ��, then the model becomes
The logit model, which assumes a logistic distribution for the error term, can be utilized to analyze the utility difference between online and in-store shopping Alternatively, a probit model may also be specified In this context, the variable reflects the utility difference and typically encompasses the characteristics of the decision-makers It is essential to incorporate alternative-specific attributes, particularly the differences between online and in-store options, such as pricing variations.
Recent studies have employed binary choice models to analyze consumer preferences for shopping channels Degeratu et al (2000) investigated how brand name, price, and information availability influence consumer behavior in online versus supermarket settings Chocarro et al (2013) assessed the effects of key purchase situation variables on the choice between traditional outlets and online stores through a survey conducted in Pamplona, Spain Crocco et al (2013) conducted an online survey in Cosenza and Rende, Italy, to identify factors affecting the preference for online shopping over in-store experiences Suel et al (2015) focused on the grocery shopping habits of London residents, exploring the relationships between online and in-store shopping Zhai et al (2016) analyzed the interplay between e-shopping and physical store shopping in Northern California, examining four stages of the shopping process for search goods (books) and experience goods (clothing) Furthermore, Arce-Urriza et al (2017) highlighted the varying impacts of price promotions on brand choice for orange juice in both online and offline channels.
In general, most of these studies conducted survey to collect data Degeratu et al.
A longitudinal experiment by (2000) explored the behaviors of online and in-store shoppers, while Croco et al (2013) analyzed respondents' last purchases, focusing on their chosen shopping channel and product categories, alongside socio-demographic factors such as gender, age, education, and income, as well as consumer attitudes measured on a five-point scale Similarly, Zhai et al (2016) examined four key variables through 42 attitudinal statements using a five-point scale Additionally, Arce-Urriza et al (2017) utilized data from past purchases made over six months in 2007 from a supermarket chain in Spain.
Logistic regression, commonly used in econometrics, enables the analysis of the multivariate effects of independent variables on an agent's choice between two alternatives, such as shopping online or in-store Key explanatory variables typically include factors that influence these shopping decisions.
Consumers' perceptions and attitudes towards online shopping are significantly influenced by the associated risks, such as credit card security concerns However, the advantages of online shopping, including real-time assistance, access to discounts, a wide variety of products, and comprehensive information, enhance the overall shopping experience Key elements like service quality and trust in online transactions play a crucial role in shaping consumer confidence in e-commerce (Chang et al., 2005; Chocarro et al., 2013; Crocco et al., 2013).
Consumer characteristics encompass socio-demographic variables, knowledge of internet and computer usage, and psychological factors, as highlighted in studies by Chang et al (2005), Chocarro et al (2013), Crocco et al (2013), and Zhai et al (2016).
The characteristics of a product or website typically include factors such as cost, trial availability, product tangibility, and design features, which are essential for comparing different products effectively (Chang et al., 2005; Chocarro et al., 2013; Crocco et al., 2013; Zhai et al., 2016).
Research indicates that online shoppers typically have less time for in-store visits but more disposable income, making them less sensitive to prices and more selective about searching for vouchers (Degeratu et al., 2000) Additionally, offline shopping channels show greater elasticity to promotions compared to online options, suggesting consumers are more responsive to in-store promotions (Arce-Urriza et al., 2017) The value of time significantly influences online shopping preferences, with factors such as distance to stores and time pressure heavily affecting the likelihood of choosing online options (Chocarro et al., 2013) Shoppers tend to favor online purchases for search goods, while demographic factors primarily influence the buying decisions for high-involvement products Furthermore, items like books are more frequently purchased online, whereas experience goods such as clothing often lead consumers to visit physical stores, as many seek information and the opportunity to try products before buying.
According to Croco et al (2013), interactions with shop assistants and concerns about credit card risks negatively influence online shopping behaviors Consumers who prioritize these interactions and risks tend to avoid online shopping Conversely, the ability to negotiate positively affects online shopping choices, indicating that individuals who value negotiation are more inclined to shop online Additionally, the immediate availability of products further enhances the appeal of online shopping.
10 has a negative impact on online shopping, meaning that if the consumers concern about the immediately having products attribute is important, they are less likely to shop online.
Frequency of online shopping
The key feature of this analysis is the dependent variable, which can take two forms The first form utilizes an ordinal scale to categorize behaviors, such as online searching frequencies (e.g., never, infrequent, frequent) and online buying habits (e.g., never, infrequent, frequent) The second form involves discrete variables that measure shopping trip frequencies, such as the average number of trips in the past week or month When the mean values of these variables are low, a Poisson model is appropriate, indicating that the data adheres to a Poisson distribution.
More specifically, the general function type for these dependent variables is usually expressed in two forms The first one is ordinal probit model, for example:
Let � = 1, 2, 3, 4 be the frequencies of online searching or online buying.
Let � ∗ a continuous latent variable of online searching or online buying frequency � ∗ can not observe, but we can observe the categories of frequency instead
4 �� � ∗ > � 3 The � coefficient and critical values � 1 , � 2 , � 3 will be estimated by the model Then theprobabilities will be calculated as below
The second model is Poisson distribution with a random variable � is said to have a
Poisson distribution with parameter � > 0 if it takes integer values � = 1, 2, 3, … ,
� is the numbers of online shopping time or frequency of online searching in the past 12 months intervalwith probability
�! where � = 2.71828 be natural logarithms, � is the average number of online searching or online buying in the past 12 months, and � is specified as a function of explanatory variables,
The frequency of online searches, online purchases, and in-store shopping is defined as a key parameter in understanding consumer behavior (Farag et al., 2005; 2006; 2007; Zhou and Wang, 2014) This model includes individual characteristics such as socio-demographics, behavioral patterns, and attitudinal variables, which are essential for analyzing shopping habits (Farag et al., 2005; 2006; 2007; Zhou and Wang, 2014).
Numerous studies have explored the patterns of online shopping and its relationship with in-store purchases Notably, Farag et al (2005) investigated the correlation between online searching frequency, online purchasing, and non-daily shopping trips, while controlling for socio-demographic, land use, behavioral, and attitudinal factors Their research utilized empirical data from Minneapolis, USA, and Utrecht, Netherlands (Farag et al., 2006) In 2007, they further examined the interplay between online buying frequency, online searching, and non-daily shopping trips, factoring in attitudes and land use characteristics Additionally, Dijst et al (2008) analyzed how demographic factors, land use, technology ownership, internet skills, and personality traits influenced online buying frequency Farag et al (2005) gathered data from 826 respondents in Utrecht and its suburbs through a shopping survey and a two-day travel diary, while Farag et al (2006) categorized participants based on their online purchasing behavior.
In the multivariate analysis, we examined the frequency of online purchases among individuals who have previously shopped online Additionally, we measured in-store shopping by calculating the average number of trips made per week for daily shopping and per month for non-daily shopping, along with the average duration of each shopping trip in minutes.
The econometrics models for this approach include Poisson model and the like (i.e negative binomial), Ordered Logit/Probit, and interval regression.
Research shows that online searching positively influences both in-store shopping and online purchases, indicating that online and in-store shopping are complementary rather than substitutes Additionally, increased Internet usage correlates with more frequent shopping trips Zhou and Wang (2014) highlight that online shopping encourages shopping trips, while shopping trips do not hinder the likelihood of online shopping.
Random Utility Models (RUM)
In this decision-making method, decision makers select the alternative that offers the highest utility from a set of multiple options (Marschak, 1960; McFadden, 1974) The utility level for each alternative is influenced by its attributes, known as the systematic component, while the random component, which remains unobservable to researchers, also contributes to the overall utility Consequently, the utility derived from an alternative encompasses both the systematic and random components (Train, 2002).
� � is usually assumed to be a function of attributes � of alternative �
� � = � 0,� + � 1 � 1� + � 2 � 2� + ⋯ + � � � �� in the context of analyzing the choice between in-store and online shopping, � � is utility from online shopping and in-store shopping alternatives, �� marginal utility of attribute
�, � �� level of attribute � of alternative �, which often contains the store’s attributes such as purchase price, travel cost, travel time, delivery cost, delivery time (Hsiao, 2009 and Schmid et al., 2016).
Under the assumption that the error component follows Gumbel distribution, the choice probability that the person chooses either store shopping or e-shopping is defined as
Few studies utilize Random Utility Model (RUM) to examine shopping mode choices due to data collection challenges Notably, Lee and Tan (2003) applied RUM to investigate consumer preferences between online and in-store shopping, identifying retail reputation and product risk as significant influencing factors Hsiao (2009) further analyzed consumer decisions between physical stores and e-shopping, emphasizing the impact of travel time and product delivery time Additionally, Schmid et al (2016) conducted a study with 339 participants in Zurich, Switzerland, evaluating the trade-offs between online and in-store shopping for experience and search goods.
Most of the studies using RUM in analysis conducted experimental design to survey the stated preference data based on revealed preference data.
Lee and Tan (2003) utilized conjoint analysis to demonstrate that experienced shoppers tend to favor low-risk purchases online Additionally, consumers show a preference for well-known brands over lesser-known alternatives when shopping in online stores Hsiao also employed conjoint analysis to explore similar consumer behavior patterns.
(2009) found that value of travel time is $5.29/hour, while value of delivery time is
$0.53/day This implies an online bookstore will have to lower a value of book’s price by
Delaying delivery by one day costs $0.53 in lost sales at physical bookstores, a significantly lower value compared to the impact of spending one hour traveling to the store Research by Schmid et al (2016) highlights that in Switzerland, the value of saving travel time is considerably higher at 40 CHF per hour, compared to just 16 CHF per hour for saving delivery time, indicating a clear advantage for online shopping.
Research indicates that high-income shoppers tend to have a favorable attitude towards online shopping (Schmid et al., 2016) Additionally, Hsiao (2009) discovered that individuals with prior e-shopping experience are more inclined to shop online to save on travel time and expenses Interestingly, even those without e-shopping experience are increasingly considering online purchases.
When travel costs rise and delivery times shorten, consumers can benefit from online book purchases A round trip to a bookstore costs approximately USD 5.58, with an average wait time of 5.48 days for a purchased book, which has a monetary value of around USD 2.90 Thus, buying a book online saves consumers USD 5.58 in travel expenses, although they incur a waiting period of about 5.48 days, equating to a cost of USD 2.90.
Structural Equation Models (SEM)
Structural Equation Modeling (SEM) is widely utilized to analyze consumer shopping channel preferences due to the ease of gathering data on dependent variables In this approach, frequency of online shopping is measured using a Likert scale, where responses range from "1" for never to "4" for very frequent This method allows researchers to effectively assess how often consumers engage in online shopping The general equation of SEM facilitates this analysis.
• � contains endogenous variables such as online shopping frequency.
• � includes exogenous variables, which normally include demographics, Internet experience and shopping attitudes.
• � coefficients representing direct effects of endogenous variables on other endogenous variables.
• � coefficients representing direct effects of exogenous variables on endogenous variables.
Numerous studies have examined the prevalence of online shopping, often utilizing the Structural Equation Modeling (SEM) method For instance, Cao et al (2012) analyzed the connections between online purchasing, traditional in-store shopping, and the process of searching for product information online.
539 adults Internet users in the Minneapolis-St Paul metropolitan area In addition, Cao et al.
(2013) used the data from the shopping survey in the Twin Cities of 585 adults Internet users to investigate the association between spatial attributes and online shopping frequency.
Zhou and Wang (2014) investigated the impact of online shopping on in-store shopping trips, focusing on whether the convenience of door-to-door delivery services diminishes the need for physical retail visits Their analysis utilized data from the 2009 National Household Travel Survey (NHTS).
The study utilized maximum likelihood estimation with a Poisson distribution, as outlined by Cao et al (2012), who employed Structural Equation Modeling (SEM) to analyze shopping behavior through variables such as online purchase frequency, in-store visit frequency, and product information search frequency They assessed shopping attitudes using a five-point Likert scale, alongside considering internet experiences In a subsequent study, Cao et al (2013) tested two hypotheses using SEM The first hypothesis examined the relationship between endogenous variables—urban area indicators and e-shopping frequency—and exogenous variables including social demographics, internet experience, and shopping attitudes, employing weighted least squares with robust standard errors The second hypothesis investigated the impact of shopping accessibility on online shopping frequency across urban, suburban, and exurban neighborhoods, articulated through three distinct equations.
Cao et al (2012) discovered that increased online shopping frequency positively influences in-store shopping, while online search frequency also enhances both online and in-store shopping activities This suggests that online shopping complements in-store experiences, although it may challenge traditional shopping habits Zhou and Wang (2014) further noted that online purchases can drive in-store shopping trips, although they also indicated that shopping trips may reduce the inclination toward online shopping Demographic, regional, and household factors significantly affect both online and in-store shopping frequencies Additionally, Cao et al (2013) found that geographical location plays a crucial role in e-shopping, as shopping accessibility varies in metropolitan areas, with urban regions exhibiting higher Internet accessibility, thus promoting more online shopping compared to areas with lower accessibility Conversely, exurban areas with limited Internet access do not encourage e-shopping.
RESEARCH METHODOLOGY
The choice of online shopping versus traditional store shopping
Random utility theory stem from the Law of Comparative Judgment proposed by Thurstone
In 1927, significant research on psychophysical discrimination explored dominance judgments between pairs of alternatives (Adamowicz et al., 1998) This research proposed that consumers, when faced with mutually exclusive choices, would select the alternative that offers the highest stimulus level The stimulus consists of both systematic and random components, represented as utility When the perceived stimulus is equated with utility, this theory transitions into an economic framework (McFadden, 2001).
(1960) has developed a theoretical framework for the choice probabilities of utility maximization that included random components, which is called Random Utility Maximization (RUM) (McFadden, 2001).
According to McFadden (2001), utility function is specified as �(�), where � is thevector of consumed quantities of various goods If the consumer prefers a bundle of goods
) The utility maximization subject to a budget constraint �� ≤ �, where � is consumer’s income and � is the vector of price with demand function � = � ,(((((((((((((( ( �) The general problem was established as following
By understanding each consumer's utility function and income, we can effectively address their individual needs This analysis enables us to determine the overall market demand across multiple consumers, represented as D = {D1, …, Dn}, where D signifies the demand for goods ranging from 1 to n.
In a discrete choice framework, the utility derived by a decision-maker from a set of alternative options is crucial for understanding consumer behavior Each consumer faces a choice set comprising multiple alternatives, and they must select only one option This selection process is represented by a binary variable for each alternative, indicating the consumer's choice.
� �� The discrete model can be operationalized by specifying the utility function, usually in a linear form where the utility of consumer � from alternative �
Ben-Akiva and Lerman (1985) proposed that the utility of each alternative is influenced by specific attributes, which play a crucial role in consumer choice behavior regarding shopping modes Key attributes affecting the selection of shopping channels include purchase price, time, and cost (Bateman et al., 2002).
Purchase price significantly influences store choice, with online retailers often offering lower prices than traditional stores due to reduced overhead costs from not needing sales staff or physical space However, online sellers must account for shipping fees, which can affect the final price Research indicates that urban consumers frequently seek online information about product prices before visiting physical stores to compare costs and gather additional information This behavior highlights the importance of understanding price disparities between online and in-store shopping channels.
The transparency of price information on the Internet significantly reduces the price discrepancies that marketers can create According to Koyuncu and Bhattacharya (2004), consumers are more inclined to shop online because this channel typically offers better prices, leading to an increase in online purchases.
Travel significantly differentiates online shopping from traditional in-store experiences, as consumers incur both travel time and costs when visiting physical stores (Hsiao, 2009) Conversely, online shoppers invest time in placing orders, await product delivery, and may incur shipping fees According to Truong and Hensher (1985), travel time can lead to disutility for individuals Hsiao (2009) further emphasizes that wasted travel time and costs negatively impact consumer value, resulting in diminished utility Therefore, for many consumers, shopping online often emerges as the more advantageous option.
After completing an online purchase, consumers typically face a waiting period for product delivery, except for digital items such as music files, software, and certain services like internet banking and online consultations that require no delivery time Delays in delivery can significantly diminish customer satisfaction and create uncertainty regarding product quality, as highlighted by Liu and Wei (2003) Additionally, research by Koyuncu and Bhattacharya (2004) indicates that longer delivery times negatively impact consumer purchasing decisions, leading to a decrease in online purchases.
Besides the main effects of these attributes, the individual characteristics and attitudinal forward shopping channels play an important role According to Adamowicz et al.
In 1998, it was established that individual choices can systematically differ, prompting the need to expand the set of explanatory variables to encompass demographic and psychological factors These individual differences are believed to impact utility levels through variations in the intercept and slope coefficients within the coefficient vector.
Customers decide between online and in-store shopping based on which option provides greater utility The utility functions for both shopping methods are influenced by various attributes.
This is the Basic model The main attributes utilized in this study is adopted from Hsiao (2009) and Schmid et al (2016).
In the equation (1), � �� is the indirect utility function of online shopping channel, which covers the key variables as
• ���������������� �� is the alternative specific constant of online shopping.
• ����� �� : The effect of purchase price of online shopping on online shopping utility.
• ��������� �� : The effect of ordering time on online shopping utility.
• ������������ �� : The effect of delivery time on online shopping utility.
• �ℎ��������� �� : The effect of shipping fee (or delivery cost) on online shopping utility.
In the equation (2), � �� is the indirect utility function of in-store shopping channel, which contains the key variables as
• ����� �� : The effect of in-store purchase price on utility of in-store shopping.
• �ℎ���������� �� : The effect of shopping time on in-store shopping utility.
• ���������� �� : The effect of travel time (to store) on in-store shopping utility.
• ���������� �� : The effect of travel cost on in-store shopping utility. The Full model in this study is estimated with attributes interacted with individual characteristics:
• Online ASC (���������������� �� ) is interacted with dummy variable for office worker, and internet access frequency (hours/day) This is to allow for difference in online
20 shopping among office workers and non-office workers, and among those with different internet access frequency.
• Prices of both alternatives (online and in-store) are interated with income. This is to allow for different price sensitivity among shoppers with different income.
• Delivery time is interacted with gender to allow for female shoppers may be more patient in waiting for the delivery.
• Travel time (to store) is interacted with income, as shoppers with higher income may have higher opportunity cost of time.
The utility function of the Full model is then:
The comprehensive model enhances understanding of consumer decision-making in selecting specific alternatives for utility maximization Beyond the primary attributes of the basic model, it incorporates additional variables as outlined in equations (3) and (4).
• ���������������� �� × ������: The additional effect of the online shoppers who are office workers and non-office workers on utility of online shopping.
• ���������������� �� × ����������: The additional effect of the online shoppers who have the internet access frequency on utility of online shopping.
• ����� �� × ������: People who have the different levels of income with respect to the sensitivity of online purchase price, which have the additional effect on utility of online shopping.
• ������������ �� × ������: People who are female or male with respect to the sensitivity of delivery time, which have the additional effect on utility of online shopping.
• ����� �� × ������: People who have the different levels of income with respect to the sensitivity of in-store purchase price, which have the additional effect on utility of online shopping.
• ���������� �� × ������: People who have the different levels of income with respect to the sensitivity of travel time, which have the additional effect on utility of online shopping.
From the estimated coefficients, we can calculate the willingness to pay (WTP) of time and cost attributes WTPs can be calculated from the Basic model as follows
In our analysis, we utilize the coefficient of price from the in-store utility function as the marginal utility of money, as immediate payment in physical stores contrasts with the delayed payment option available for online shopping This distinction suggests that the online shopping utility function may not accurately represent the marginal utility of instant money We present several equations to quantify willingness to pay (WTP): Equation (5) calculates the WTP for one minute spent searching and ordering online (VND/min), while Equation (6) assesses the WTP for one minute spent finding and purchasing items in-store (VND/min) Additionally, Equation (7) quantifies the WTP for one day of waiting for product delivery (VND/day), and Equation (8) determines the WTP for one minute of travel to a store (VND/min).
Table 3 1: The store choice attributes description
Variable Unit Description Expected sign
����� �� VND 1000 Purchase price of online shopping: the price when buying online.
��������� �� Minute Ordering time: the time spends for searching and placing the order of books over the internet.
������������ �� Days Delivery time: the day of waiting for the delivery of purchased books
��� = (5) ��� ��� = (6) ��� (within 1 day, 2-3 days and >= 3 days).
�ℎ��������� �� VND 1000 Shipping fee (or delivery cost): the monetary value which pays for the delivery of purchased books (VND 0, VND 15 and VND 22).
����� �� VND 1000 Purchase price of in-store shopping:
The purchase price at the bookstore.
�ℎ���������� �� Minute Shopping time: the time spent for finding and buying books at the bookstore.
���������� �� Minute Travel time: the time spends to go to the bookstore.
���������� �� VND 1000 Travel cost: the monetary value which depends on the reported transportation and the distance to the store for the last purchase in the questionnaire.
Calcul ating the WTPs using the Full model is more complicated
In the Full model, the marginal utility of money is � 1 + � 1 �̅̅�̅�̅̅�̅̅�̅̅
�̅̅�̅�̅̅�̅̅�̅̅�̅ is an average income The WTPs for these attributes is now
In the same manner, the WTPs of these attributes in equation
(9), (10), (11) and (12) are also described as in the Basic model
However, the marginal utility of delivery time is
The choice of shopping mode � ∈ {��, � } follows the utility maximization for respondent � is modeled as
Therefore, the probability of choosing shopping channel is calculated as
The study outlines the definitions and units of measurement, along with the anticipated signs of alternative-specific attributes and socio-demographic variables, as detailed in Tables 3.1 and 3.2.
This study utilizes both Revealed Preference (RP) and Stated Preference (SP) data, organized in two stages for data collection Stage I involves questionnaires based on the RP method to gather revealed data from respondents Stage II features a choice experiment, or SP method, developed from the insights gained in Stage I, allowing for the collection of stated choices from participants, as outlined by Louviere et al.
The revealed preference (RP) method gathers data on actual shopping behaviors, reflecting "the world as it is," while the stated preference (SP) method explores potential shopping choices, illustrating "the world as it could be." Essentially, RP surveys capture what respondents have actually done, whereas SP surveys project what respondents might do in hypothetical situations.
3 g o v e rn m en t o ffi ce rs; 4 fr ee la nc er s; 5 co m pa ny e m pl oy ee s; 6 re se ar c h e r s
In hours 1 = 2 – under 4; 2 = 4 – under 6; 3 = 6 – under 8; 4
Figure 3.1 describes the conceptual framework of the choice between online shopping versus in-store shopping.
Revealed preference data: methods of collection
Revealed Preference Theory, introduced by Nobel laureate Paul Samuelson in 1938, seeks to uncover individual preferences or utility functions through the observation of actual choice behavior A key benefit of revealed preference data is that it reflects genuine decisions made by respondents in real-life situations, as noted by Morikawa in 1989.
The revealed preference technique effectively addresses issues associated with hypothetical responses, which often fail to accurately reflect choice behavior and strategic responses (McFadden, 1974) In utility-maximizing decision-making, agents compare alternatives by assessing benefits derived from actual observed risks (McFadden, 1974a) Furthermore, Louviere et al (2000) highlighted that revealed preference is a suitable method for analyzing a small number of observations without requiring specific functional form assumptions.
The application of revealed preference relies on observed choices, making it difficult to ascertain the true choice set Ben-Akiva and Lerman (1985) noted that revealed preference surveys yield fewer observations due to high costs Additionally, the strong correlation among attributes like travel time value, delivery time value, and travel cost can lead to biased estimates of the effects of independent variables on utility The limited range of attributes stems from their dependence on the specific decisions made in existing product consumption scenarios (Adamowicz et al.).
Multicollinearity, estimation efficiency, and endogeneity are potential issues highlighted by Adamowicz et al (1994), who argue that revealed preference (RP) methods are more effective in estimating use and market value However, RP data often struggle to capture intangible attributes like reliability, convenience, and comfort Furthermore, RP surveys are not suitable for assessing demand for new modes of transportation due to the lack of available information or existing alternatives; instead, stated preference surveys are recommended for this purpose (Morikawa, 1989; Ben-Akiva et al., 1994; Louviere et al.).
2000) Hence, the preference indicator is usually as choice (Morikawa, 1989).
The application of revealed preference relies on observed choices, making it difficult to accurately determine the choice set Ben-Akiva and Lerman (1985) noted that revealed preference surveys yield fewer observations due to high costs Additionally, the strong correlation among attributes such as travel time value, delivery time value, and travel cost may lead to biased estimates of independent variables' effects on utility Furthermore, the limited range of attributes is influenced by the specific decisions made in existing product consumption scenarios (Adamowicz et al.).
In 1997, issues such as multicollinearity, estimation efficiency, and endogeneity were identified as significant challenges in research Adamowicz et al (1994) argued that revealed preference (RP) methods provide more accurate estimations of use and market value; however, measuring intangible attributes like reliability, convenience, and comfort remains difficult for researchers Additionally, RP surveys are not suitable for estimating demand for new attributes due to the lack of available information, making stated preference surveys a more appropriate alternative (Morikawa, 1989; Ben-Akiva et al., 1994; Louviere et al., 2000).
A personal interview was conducted to explore respondents' shopping mode choices, utilizing a questionnaire to capture real-life responses Building on previous research, we hypothesize that varying shopping modes offer distinct utility levels based on shoppers' intrinsic attributes This study aims to derive the utility function for shopping channels—both online and in-store—by analyzing choice decisions The questionnaire specifically targets respondents' preferences for shopping modes related to their book purchases over the past year.
To determine the suitability of respondents for the interview, they will first be asked if they purchased a book in-store or online in the past 12 months If they answer yes, the next question will be whether they were the decision maker for that last purchase Only those who confirm they were the decision maker will continue with the interview.
The study gathered general information from respondents regarding their book purchasing habits over the past year by asking about the frequency of purchases and the recency of their last transaction Additionally, participants were prompted to share their past experiences with book shopping and internet usage Key questions included identifying the most commonly used devices for internet access—such as mobile phones, laptops, PCs, or tablets—and the average daily hours spent online This information aimed to categorize individuals based on their history of online book purchases.
“Did you ever use the Internet for purchasing books”.
To understand the motivations behind book purchases, we asked respondents why they decided to buy their latest book Additionally, we inquired about the location of their most recent purchase, whether it was through an online channel or in-store This information allowed us to further explore their experiences with both online and in-store shopping for books.
Respondents who shopped in-store were asked to answer questions about online shopping based on their knowledge, and vice versa (Hsiao, 2009) To analyze preferences, we classified attributes related to their most recent purchase in both channels, specifically inquiring about the total amount spent on their last book purchase (in VND).
When evaluating your most recent purchase, consider whether the total price was higher or lower compared to previous transactions It's important to quantify this change, so determining the percentage difference can provide valuable insights into your spending habits.
To gather valuable insights on shopping and ordering time, we will ask respondents several key questions, including the total minutes spent on shopping and ordering Additionally, we will inquire if their last purchase involved only books and how many minutes were dedicated to travel time for that trip We will also explore delivery time by asking how many days it took to receive their order To assess travel costs, we will ask about the total expense incurred for the last purchase, and if respondents are unsure, we will suggest questions regarding the mode of transport used and the distance to the bookstore Lastly, for online purchases, we will inquire about the delivery cost associated with their last order.
Each participant provided personal information, including gender, age, marital status, education, occupation, and income level Additionally, they rated 18 attitudinal statements using a five-point Likert scale This data collection forms a foundational database for stated preference research.
The stated preference data: methods of collection including choice experimental design 29
choice experimental design 3.3.1 The stated preference method
The stated preference method utilizes hypothetical scenarios to gauge respondents' choices, contrasting with their actual behaviors in real-world situations This approach allows for a greater number of observations in experiments, as the researcher controls the process of decision-making for the respondents.
Morikawa (1989) highlighted the advantages of stated preference methods, noting that researchers can clearly define the choice set presented to respondents This approach allows for an expanded range of attribute levels, independent of real-world variable constraints Additionally, it minimizes high collinearity issues, such as those between time and cost, by enabling researchers to thoughtfully combine attribute levels during experimental design, thus reducing measurement errors The method also facilitates the collection of intangible variables through respondent ratings, allowing for the inference of preferences for new alternatives As a result, preference indicators can be effectively elicited through various means, including choice, rating, and ranking, indicating that data gathered from stated preference surveys often surpasses that of revealed preferences.
Stated preference data often lacks the validity of revealed preferences, as highlighted by Morikawa (1989) This is primarily due to the reliance on hypothetical scenarios presented to respondents, which can lead to biases in their choices and estimations (Morikawa, 1989; Fifer et al., 2014).
Table 3.3: Attributes and levels in stated choice experimental design
Delivery time (OL) Days Within 1 day, 2 – 3 days, > 3 days
Each attribute consists of various levels, and by analyzing the responses from stage I, we can uncover the preferences that inform these levels This process aims to improve the revelation of individual preferences (Rose et al., 2008; Schmid et al., 2016) Table 3.3 illustrates the hypothesized attributes and their corresponding levels that are expected to influence the decision-making process between online and in-store shopping.
Based on Table 3.3, the pivot design produces the different individual choice situations. The attribute’s levels were modified depending on the reference value what we collected from revealed preference data.
To effectively create stated choice questionnaires, researchers must carefully construct alternative scenarios for respondents (Louviere et al., 2000; Bateman et al., 2002) By utilizing various combinations of attribute levels, profiles are generated through orthogonal design methods (Louviere, 1988; Adamowicz et al., 1998).
We code the variables of in-store shopping as purchase price � 1 , shopping time
In a study analyzing travel options, four attributes were identified, with travel time having four levels and travel cost, along with two other attributes, having three levels each This resulted in a total of 108 possible profiles through a full factorial design, characterized by orthogonality, meaning there is no correlation between any two variables However, the sheer number of alternatives generated makes it impractical to present all 108 profiles to respondents simultaneously, as it would be unrealistic to expect them to evaluate such a large set of options at once.
Experimental design methods are created to address the challenges associated with complete factorial design by selecting a sample of profiles that possess specific statistical properties, enabling the estimation of utility specifications (Adamowicz et al.).
In 1998, a study utilized a subset of full factorial design to enhance the statistical efficiency of a questionnaire (Bateman et al., 2002) By employing a fractional factorial design, the research effectively reduced the number of scenario combinations while maintaining the integrity of the existing data.
47 or some interactions will not be detected However, fractional factorial design becomes effective in practical when presenting to respondents.
In experimental design, we can estimate two types of effects of attributes on choices: main effects and interaction effects (Bateman et al., 2002) Main effects reflect the impact of individual personal attributes and are derived from the orthogonal subset of the full factorial design, allowing researchers to assess the influence of independent variables on the values of other variables (Adamowicz et al., 1998) Conversely, interaction effects reveal the relationship between behavior and the variations in combinations of different attributes (Bateman et al., 2002), illustrating how the effect of one variable depends on the values of others.
This study employs a choice experimental design utilizing D-optimal designs, incorporating both linear and quadratic functions Each attribute is systematically coded with numerical values ranging from one to four, as outlined in Table 3.4.
Constructing choice sets is essential because, after experimental design generates profiles, these profiles are organized into choice sets for respondents (Bateman et al., 2002) Choice sets can be formed by combining two or more alternatives In this study, the choice sets consist of two alternatives: online shopping and in-store shopping When respondents select between these shopping channels, they engage in choice scenarios or tasks.
Table 3.4: The attribute’s levels of online and physical store
(OL) (IS) (OL) (IS) (OL) (IS) (OL) (IS)
To effectively present alternatives to respondents, it is crucial to bundle them thoughtfully A common approach to constructing paired choices involves highlighting the differences between each option (Bateman et al., 2002) For instance, in online shopping, one might assign a purchase price value of 0%, with ordering time at -50%, free delivery, and a delivery time of one day In contrast, in-store shopping may have a similar or different purchase price, with varying shopping times It's essential to address situations in the questionnaire where dominance needs to be eliminated, relying on either the reference case of interest or the researcher’s prior experience.
Dividing profiles into manageable blocks serves the purpose of simplifying the design process for respondents by limiting their choice scenarios This approach ensures that participants are not overwhelmed, as they cannot complete all tasks at once Ideally, the blocks are randomly assigned while taking into account the dominance of choices within each block and comparing them to subsequent blocks, allowing for a balanced and effective data collection process.
The choice experimental design in this thesis is presented in Table 3.5 The design has
The study utilized D-optimal Design to create 24 choice sets, organized into 4 blocks, with each block containing 6 choice tasks Each task presents respondents with a choice between one online shopping alternative and one in-store shopping alternative Participants are required to complete six choice tasks from a single block, making their selection between the two options provided.
Table 3.6 presents a simulated question regarding consumer preferences between online shopping and in-store purchases Utilizing revealed preference data from the respondents' last order, we populate the choice set with relevant figures For instance, if the last purchase amounted to 100,000 VND, the online shopping alternative is priced at 50,000 VND, reflecting a 50% reduction, while the in-store price remains unchanged at 100,000 VND.
Sample size and sampling
When conducting data collection, it is crucial to balance the costs involved with the desired sample sizes necessary for accurate predictions This thesis emphasizes a choice model that considers individual differences, allowing for the specification of minimum sample sizes to improve prediction accuracy within specific segments Additionally, the sample size required for stated choice experiments is influenced by the total number of hypothetical choice situations and the variety of alternatives presented in each scenario (Adamowicz et al., 1998).
• t: number of choice tasks answered by each respondent (6).
• a: number of alternatives in each choice task (2)
• c: the maximum number of levels of attributes (4 in this study)
The study required a minimum of 167 respondents for the survey, ultimately surveying 352 book buyers After excluding incomplete questionnaires, the final sample comprised 321 respondents Each participant answered six choice questions, resulting in a total of 1,926 choice observations.
Survey
To effectively select a sample frame population, researchers must determine an appropriate sampling design and survey mode, as different modes impact access to individuals and data collection costs (Bateman et al., 2002) In this study, the sample frame includes random individuals visiting the bookstore on Nguyen Van Binh book street, as well as students and staff from the School of Economics, HCMC, Ho Chi Minh City University of Pedagogy, University of Social Sciences and Humanities, and patrons of the Book Café Store and nearby parks in HCMC between July 4, 2017, and August 11, 2017 Eligible respondents are individuals over 18 years old with a source of income, and the chosen survey method is face-to-face interviews using paper-and-pencil questionnaires administered by enumerators.
The survey participants were randomly selected and first asked about their book purchasing frequency and the timing of their last purchase within the past year Only those who were decision-makers for their last book purchase proceeded to the next sections of the questionnaire, which gathered demographic information about the decision-makers Respondents then provided insights into their past experiences with both online and in-store shopping, and evaluated the importance of various shopping process factors using a five-point Likert scale Additionally, each participant specified their purpose for buying books and indicated whether their last purchase was made online or in-store.
In the second stage, respondents are required to disclose essential details about their most recent purchase, whether made online or in-store This includes information on the purchase price, the time spent shopping, and travel costs If the last purchase was made online, respondents will also be asked to provide similar details as if they had shopped in-store.
In the final stage of the study, respondents select their preferred shopping mode from two alternatives based on varying attribute levels This process is known as a stated choice experiment.
Participants will be evaluated based on their ability to answer all questions accurately The design features a combination of four distinct blocks, with each respondent randomly assigned to one block containing six hypothetical scenarios.
In this thesis, a total of 352 respondents participated in the survey, with each block being answered by a minimum of 88 respondents To encourage participation, each respondent who completed the full questionnaire received an incentive of VND 50,000 during the interviews.
The questionnaire
The questionnaire is divided into four sections: Section 1 gathers general information on shopping habits, experiences, demographics, and attitudes towards online and traditional shopping Sections 2 and 3 focus on revealed preferences from the most recent shopping experience, comparing online and in-store purchases Finally, Section 4 features simulated choice scenarios For more details, refer to the Appendix containing the complete questionnaire.
Estimation methods
The estimation method aims to identify and estimate the parameters related to variables that explain the shopping choice model between online and in-store channels The ultimate goal is to obtain unbiased estimates of the parameter vector, which encompasses the marginal utilities of various attributes (Adamowicz et al., 1998) By maximizing the log-likelihood function, a conditional logit model can effectively estimate these parameters.
The conditional logit model estimates parameters that are assumed to be constant for all respondents in a sample However, if there is heterogeneity in preferences among respondents, the basic assumption of the simple conditional logit model becomes invalid.
The Mixed Logit model is the ideal econometric model for estimating the taste coefficient due to its flexibility and applicability to any random utility model (McFadden and Train, 2000) This model effectively addresses challenges such as random taste variation, unrestricted replacement patterns, and unmonitored correlations over time, which are limitations of the standard logit model (Train, 2003).
The Mixed Logit model, as cited in 2003, is not confined to a normal distribution, allowing for straightforward derivation and simple calculations of choice probabilities in hypothetical scenarios Its formation can be influenced by various behavioral specifications, leading to distinct interpretations based on different behaviors Essentially, the Mixed Logit model is defined by the functional approach to its choice probabilities, where any behavioral characteristics derived from this specific form are recognized as part of the Mixed Logit framework.
Mixed logit probabilities, as described by Train (2003), are defined by integrating standard logit probabilities over a density function In essence, a mixed logit model represents any model where the choice probabilities of a decision maker can be articulated in this manner.
• � �� (�) is the logit probability evaluated at parameters �:
• � �� (�) is the observed portion of the utility If utility is linear in �, then
� �� (�) � ′� �� In this case, the mixed logit probability gets its normal form as
The mixed logit model represents a weighted average of Logit probabilities at various coefficients, where the weights are determined by a density function, denoted as \( f(\theta) \) This statistical approach is characterized by combining multiple functions, with the mixing distribution serving as the density that assigns these weights Essentially, the mixed logit formula is evaluated at distinct values of \( \theta \) in conjunction with the mixing distribution \( f(\theta) \) However, a common challenge arises in the terminology of parameters within mixed logit models, which consist of two sets of parameters: those belonging to the logit function and those associated with the density \( f(\theta) \) For instance, if \( \theta \) follows a normal distribution with mean \( \mu \) and covariance \( \Sigma \), the density function \( f(\theta) \) is represented by \( \mu \) and \( \Sigma \) Researchers typically focus on estimating the parameters of the density \( f \), as this influences the choice probability under the specified density.
McFadden and Train (2000) emphasized that the careful selection of variables and the appropriate mixing distribution for mixed logit models can achieve a high degree of accuracy in approximating any Random Utility Model (RUM) This approach represents the most straightforward derivation and is widely utilized in the field.
∑ recent applications, in which is based on random parameters The utility of consumer � from alternative � ∈ {��, �� } is specified as ′ � + �
Where � ′ is a vector of coefficients that depict for these variables of consumer � representing the consumer’s preference Random component � �,� has IID extreme value.
Because, the coefficients of decision makers are varied in the population with density
� � coefficients are what the researcher want to know and estimate on �′.
Therefore, taking the integral of � �� (�′ � ) over all possible variables of
� � unconditional choice probability: that is
The researcher can specify a distribution for coefficients and estimates the parameters of that distribution.
RESEARCH RESULTS
Summary statistics
The survey approached 352 individuals who are decision makers in the last purchase within
The survey, conducted between July 4 and August 11, 2017, initially included responses from a diverse group; however, 31 questionnaires were excluded due to incomplete answers or respondents' inability to recall their last purchase, particularly among older individuals and those under 18 without a monthly income Consequently, the final analysis was based on 321 valid responses, summarized in Table 4.1.
The study involved 321 respondents who participated in 6 choice tasks, resulting in a total of 1,926 stated choice observations Table 4.2 presents a descriptive analysis of the sample, highlighting key socio-demographic factors such as gender, age, marital status, education, occupation, income, and parental status The gender distribution is fairly balanced, with males representing 41.74% and females 58.26% Notably, 72.9% of participants have previously purchased books, with a higher percentage of female buyers at 60.26% compared to 39.74% male buyers Respondents' ages range from 18 to 58 years, with an average age of 24.58, indicating a predominance of younger individuals in the sample.
A significant majority of respondents, 93.77%, are aged between 18 and 35 years old, with the highest propensity for online shopping observed among those aged 20 to 30, who represent 75.64% of this demographic.
Table 4.1: Survey location and number of respondents
Ho Chi Minh Book Street 169
University of Economics HCMC (UEH) 22
University of Social Sciences and Humanities 2
The study reveals that the respondents have a high level of education, with an average of 15.6 years of schooling, where 88.16% hold college or university degrees Nearly half of the participants are students (49.22%), followed by office workers (31.15%), while state officials, lecturers, and teachers represent 9.03% Online book buyers predominantly consist of students (48.72%) and office employees (31.2%) Financially, almost half (48.6%) earn under 5 million, and 42.99% have incomes ranging from 5 million to 15 million Additionally, a significant majority of respondents are single (85.67%) and do not have children (88.79%).
Table 4.2: Descriptive statistics of the sample
Having children Not having children
Over 10 hours Mobile phone Laptop, PC Tablet
Frequency of Internet access Percentage of Internet access devices
The statistics suggest that online book shoppers are usually female and young, being students and the officers, having monthly income in the range of 0 to 10 millions.
The data presented in Figure 4.1 illustrates the frequency of Internet access among respondents, revealing that 100% of the sample utilizes the Internet Notably, 39% access the Internet for 2 to 4 hours daily, while 25% spend 4 to 6 hours online Additionally, 11-12% of respondents use the Internet for 6 to 10 hours, and 12.5% exceed 10 hours of usage Furthermore, over half of the respondents (58.57%) prefer mobile phones for Internet access, compared to 38.94% who use laptops and PCs, and only 2.49% who utilize tablets.
Table 4.3 highlights the significance of attitudes towards online versus store shopping, based on responses from a sample of 18 issues Notably, 68.23% of respondents value immediate ownership of books after payment, with 63.24% of online book purchasers echoing this sentiment Additionally, 52.34% of all respondents and 52.13% of online book buyers emphasize the importance of having space for movement and social interaction In contrast, 59.5% of respondents and 63.25% of online buyers deem interaction with shop assistants unimportant Furthermore, a substantial 86.3% of the sample and online bookstore customers prioritize the opportunity to read books before making a purchase.
Figure 4.1: Frequency of Internet access and Internet access devices
A significant majority of respondents, 57.94%, highlighted the ease of replacing books as a crucial factor in their purchasing decisions, with 60.26% of online buyers agreeing Additionally, 65.73% of participants and 67.52% of online shoppers emphasized the importance of discount programs when buying books In contrast, the variety of products available was deemed unimportant by 59.63% of the sample Furthermore, the motivation to buy books for multiple purposes, such as commuting or spontaneous purchases, was considered insignificant by 74.14% of respondents and 76.92% of online buyers Conversely, a strong 87.54% of all participants and 88.03% of online purchasers valued the ability to check a book's quality before making a purchase.
Table 4.3: Descriptive statistics analysis of consumer attitudes
Shopping attitudes Mean Std Err.
Having space to move and communicate 3.30 056
Easiness to replace the book 3.46 059
Often have the discount program 3.70 056
Can check the quality of book before purchasing 4.16 045
Can find out rare books, old books and expired issues 3.82 054
Do not have to move too far 3.77 051
Limited contact with many people 2.77 053
Avoid traffic jams or difficulty in finding parking plot 3.55 057
The importance of gift wrapping services is minimal, with only 69.47% of respondents considering it significant In contrast, the discovery of old, rare, and expired books is deemed crucial by 70.72% of participants When it comes to pricing, 67.42% of individuals find the cost of books when comparing online and in-store shopping important Additionally, over half of the respondents (57.63%) emphasize the significance of price comparison, while a substantial 81.93% prioritize convenient payment methods.
A significant majority of online book buyers value convenience, with 73.08% citing time and travel as crucial factors Additionally, 74.79% of these consumers appreciate the ability to shop without the hassle of traveling far The avoidance of traffic congestion and parking difficulties is also important to 62.82% of online shoppers Interestingly, the limited interaction with others while shopping online is deemed unimportant by 81% of respondents, highlighting an unexpected trend in online purchasing behavior.
Table 4.4: Revealed data between online shoppers and in-store shoppers
Ordering time/ Shopping time (min) 27.4 2 360 50.4 1 240
Shipping fee/ Travel cost (VND1000) 9.54 0 250 11.34 0 150
Delivery time (days)/ Travel time (min) 3.0 1 20 20.4 2 120
Table 4.4 presents a comparative analysis of book shopping behavior between online and store shoppers, revealing that the average purchase price is similar, with online shoppers spending VND 213,574 and store shoppers VND 237,409 Online shoppers save time, averaging 27.4 minutes for ordering, compared to 50.4 minutes spent by store shoppers Additionally, the difference in costs is minimal, with online delivery costs averaging VND 9,538 and store travel costs at VND 11,339 Delivery time averages three days, while travel time to bookstores is about 20.4 minutes.
Estimation results
This section outlines the estimated utility functions derived from Random Utility Models (RUM), featuring two distinct models The first model, termed the Basic model, focuses solely on attributes, providing estimates of the marginal utility of these attributes but assumes uniformity across all respondents' utility functions In contrast, the second model, known as the Full model, incorporates interactions between attributes and individual characteristics, capturing the variability in marginal utility among respondents Both models are analyzed using the conditional logit and mixed logit approaches.
As mentioned, the two utility functions for in-store and online shopping are
In this context, the Average Selling Cost (ASC) for the in-store channel is set to zero, while the ASC for the online channel is estimated Prices are expressed in thousands of VND, calculated against the cost of the most recent purchase.
Online shopping order times are quantified in minutes, encompassing the duration spent searching for products and completing the purchase Similarly, in-store shopping times are also measured in minutes, reflecting the time customers invest in locating items before making a purchase.
Online shopping In-store shopping
Delivery time is a crucial factor for online shopping, measured in days, while for in-store shopping, the equivalent metric is travel time, which refers to the minutes customers spend traveling to the store.
When shopping online, a shipping fee is incurred, typically expressed in thousands of VND, whereas in-store shopping involves travel costs, also measured in thousands of VND, that account for the expenses related to traveling to the store.
Table 4.5 presents descriptive statistics for alternative-specific attributes related to store choice, based on 1,926 observations across shopping channels In online shopping, the average purchase price is VND 171,140, with prices ranging from a minimum of VND 9,450 to a maximum of VND 1.8 million The average shipping fee is VND 12,280, with a minimum of free shipping and a maximum of VND 22,000 The average time taken to order books online is 22.21 minutes, with the quickest order placed in just 1 minute and the longest taking up to 450 minutes On average, customers wait 1.96 days for delivery, with the fastest delivery occurring within 1 day and the longest exceeding 3 days.
Table 4.5: Summary statistics of alternative-specific attributes
Mean Min Max Mean Min Max
Ordering time(OL)/Shopping time (IS)
Delivery time(OL) (Days)/Travel time (IS)
Shipping fee (OL)/Travel cost (IS) 12.28 0 22 12.26 0 195
In-store shopping reveals an average purchase price of VND 222,090, with prices ranging from a minimum of VND 12,000 to a maximum of VND 2.3 million The average travel cost to the store is VND 12,260, while walking incurs no travel cost, and the highest travel expense reaches VND 195,000 Shoppers spend an average of 48.74 minutes searching for items, with shopping times varying from a minimum of 0.8 minutes to a maximum of 264 minutes Additionally, the average travel time to the store is 22.34 minutes, with a minimum of 1.4 minutes and a maximum of 180 minutes.
The estimation results from the basic model, presented in Table 4.6, utilize both conditional logit and mixed logit models For online shoppers, the conditional logit model reveals significant negative coefficients for purchase price, delivery time, and shipping fees, indicating their adverse impact on utility Conversely, the ordering time variable appears insignificant, suggesting it may not be a substantial concern for respondents In contrast, for in-store shoppers, while the purchase price negatively affects utility significantly, the coefficients for shopping time, travel time, and travel costs are insignificant This may imply that in-store shoppers prioritize the shopping experience over time and costs associated with traveling to the store, possibly viewing shopping as a leisurely activity or incorporating it into multi-purpose trips Consequently, the travel time and costs related to visiting bookstores hold little significance for these respondents.
The analysis reveals that the purchase price coefficients for online shopping and in-store shopping are -0.012 and -0.008, respectively This suggests that individuals who favor online shopping experience a higher unit disutility compared to those who prefer shopping in physical stores.
Increasing the purchase price by one thousand VND results in a 0.012 decrease in utility for online shoppers, while in-store shoppers experience a 0.008 decrease in utility.
The analysis reveals that an increase of one minute in delivery time results in a disutility of 0.319 for frequent online shoppers, while a rise of one thousand VND in shipping fees corresponds to a disutility of 0.022 Additionally, the positive coefficient of the online channel, at 1.213, indicates that shopping online enhances utility significantly, assuming all other factors remain constant.
Table 4.6: Estimation results of the basic model
Note: Standard errors in parentheses; significant level: *** p
