Understanding the Impact of COVID-19 Pandemic on Online Shopping and Travel Behaviour: A Structural Equation Modelling Approach

Abstract

The outbreak of the COVID-19 pandemic has led to significant alterations in people’s social and economic behaviour. This paper aims to study the pandemic’s influence on online shopping and travel behaviour and discover how these phenomena are related. To this end, eight variables were identified that describe socio-demographic status, COVID-19 variables, online shopping variables, and travel behaviour. The structural equation modelling (SEM) approach was adopted to analyse the relationships between these variables. A conceptual model was formed by devising hypothetical relationships, and then the validity and reliability of the model were evaluated using SEM tools. Among the 19 theoretical relationships, 17 were verified. It was found that socio-demographic status directly affects the COVID-19 variables, influencing online shopping variables. As a result, it was inferred that during the pandemic, people’s daily travel habits had been affected by their inclinations toward online shopping, and the more people are aware of COVID-19 and feel responsible about the pandemic, the more they are persuaded to shop online rather than in-person shopping. Policymakers can use the findings of this study to change the public’s travel and shopping behaviour to tackle the pandemic.

1. Introduction

The COVID-19 pandemic has dramatically reshaped the main pillars of human societies in every aspect, such as education, economics and trade, business, communication, and transport. Many governments placed restrictions on social and economic activities during this pandemic, including closing schools and gyms, suspending food service and social functions, and ceasing unnecessary transportation to minimize human contact [1]. Moreover, since some regulations have been set on shop opening hours or lockdowns, social media has become more popular than ever, resulting in a boom in online shopping [2]. In addition, since COVID-19 is mainly transmitted through face-to-face contact, the importance of e-commerce has been increasingly recognized in a way that many consumers prefer to fulfill their shopping needs through online services [3,4]. Moreover, the mandatory usage of face masks and social distancing caused a significant decrease in physical store shopping [5]. Global traffic on retail e-commerce websites increased from 16.07 billion worldwide in January 2020 to 22 billion in June 2020, which could be a result of the coronavirus pandemic [6].

Furthermore, since some retail enterprises may not take protective measures seriously when dealing with the COVID-19 virus, customers are increasingly inclined to shop online [7]. In addition, many people around the globe have had some type of exposure to food insecurity during the COVID-19 pandemic [8], leading to people being more aware of food safety and changing their consumption behaviour [3]; as well as a change in consumption patterns, such as a decrease in eating out, showing more disposition toward cooking at home [3], and more inclination toward buying groceries online. Therefore, people are generally less willing to shop from physical grocery stores to prevent transmitting the disease [4]. As a result, some businesses struggled, and others thrived in areas, such as online entertainment, food delivery, online shopping, and online education [5].

On the other hand, studies show that there is a two-sided relationship between travel behaviour and the COVID-19 pandemic variables regardless of travel distance or population [6]; namely, there is a meaningful lag (2 to 3 weeks) between peaks in COVID-19 infections and peaks in traffic volume. There is generally a negative correlation between public transport usage and the number of reported COVID-19 infections [7]. This indicates that the COVID-19 pandemic has resulted in intra-city travelers being less inclined to use public transportation modes and prefer to travel with private cars or non-motorized means of transport [8]. Changes in travel behaviour may also be predicated on changes in shopping behaviour due to COVID-19-related restrictions. Nevertheless, the impacts of physical shopping restrictions on traffic volume were much higher in the early stages of the pandemic compared to the succeeding COVID-19 peaks [9].

In addition to the changes in shopping and consumption behaviour, the COVID-19 pandemic has significantly changed people’s travel habits, mainly through preventive measures implemented by the government, social distancing regulations, and mandatory self-quarantine protocols [10]. Although the lockdown was temporary and has now been lifted in almost all areas, and even public transport has resumed its operations [11], the number of non-business trips has remained lower than in the pre-COVID-19 era. The popularity of remote work partly influences this behavioural pattern during the pandemic [12]. Moreover, several other factors have been found to influence travel behaviour as a result of the COVID-19 pandemic, including basic socio-demographic characteristics, vehicle ownership, level of social anxiety, risk perception, weekly trip frequency before the pandemic, and behavioural changes in response to the coronavirus [13].

Since the COVID-19 pandemic has affected both shopping and travel habits, it is critical to investigate the correlations between the COVID-19 pandemic, shopping behaviour, and travel habits. Ghodsi et al. [14] applied ISM and IAHP to identify and evaluate the impact of several (psychometric) variables from transportation experts’ perspectives. Following their study, this research intends to explore the impact of the pandemic on online shopping and travel behaviour. The conceptual model was influenced by the previous research, and it completes the analysis of the variables that were found therein. Structural equation modelling (SEM) is used in this study to assess the relationships between the factors that describe these phenomena.

The main aim of this paper is to bring insight into how the psychometric variables of online shopping behaviour and COVID-19 pandemic pertain to consumers’ travel behaviour. For this purpose, Section 2 provides a review of the theoretical background and a brief discussion about the existing research gap of the problem addressed in the paper. Section 3 describes the study area in which the variables are analysed. In Section 4, the research methodology is explained in detail and the results of employing the SEM approach are presented step-by-step. Section 5 elaborates on the outcomes of the study and discusses the obtained results. Finally, Section 6 presents conclusion remarks, research implications, and some suggestions for future studies.

2. Literature Review

The literature review in this research comprises two sections. In the first section, the theoretical background of the study is briefly discussed. In the second stage, the closely related studies are reviewed to find the research gap and explain how the present study contributes to the literature.

2.1. Theoretical Background

In 1991, Ajzen [14] developed the planned behaviour theory to predict and explain various human behaviours in specific settings and deal with arbitrary complexities of social behaviour. Then, it became the basis for thousands of studies that deal with human behaviour alterations in response to different external stimuli. According to the planned behaviour theory, three independent variables determine the intention to perform a behaviour. First, an attitude refers to a person’s opinion of what is deemed to be acceptable or unacceptable behaviour, which indicates their judgments. The second indicator is the subjective norm, which describes the influence people believe is exerted upon them to follow or not follow specific behaviour. The third predictor is perceived behavioural control, which indicates the individual’s perception of the ease or difficulty of following a type of behaviour, and concerning the individual’s experience, it reflects their responsiveness to the occurrence of impediments and obstacles that could lead to the violation of the norm.

Among the research conducted before the pandemic that studied consumer shopping behaviour, Shang and Chen [15] used the SEM method to propose a theoretical model to explain consumers’ intrinsic and extrinsic motivations for online shopping. They found that inherent incentives, such as entertainment, fashion, and cognitive absorption, are more influential in online shopping than extrinsic motivations, such as economic factors and perceived usefulness. Concerning online shopping acceptance, Ha et al. [16] studied nine groups of factors that influence online shopping. These factors include socio-demographic characteristics (gender, age, income, education, culture), Internet experience (internet apprehensiveness, frequency of internet usage, comfort with the Internet), normative beliefs, shopping orientation, shopping motivation, personal traits (innovativeness), online experience (emotion, flow), psychological perception (risk perception, benefit perception, online purchasing apprehensiveness), and online shopping experience (frequency of online purchases, satisfaction level with past online transactions). Using the conformity factor analysis and SEM, Wen and Prybutok [17] presented an integrated theoretical framework to describe the factors influencing customers’ intentions to repeat online shopping. They discovered that utilitarian factors (perceived ease of use and usefulness) and hedonic factors (perceived enjoyment) significantly influence customer online repurchase intentions. Moreover, Zhou and Wang [18] employed the SEM method to investigate the links between online shopping and shopping trips. They argued that there is a two-sided relationship between these elements. More precisely, online shopping boosts shopping trips as online shoppers travel to stores to test, compare or pick up products. On the other hand, shopping trips reduce inclinations toward online shopping. They concluded that exogenous factors (e.g., demographic status, regional-specific factors, and household attributes) influence shopping trips and online shopping.

Many studies explored the influence of the COVID-19 pandemic on human behaviour. Nindrea and Sari [19] investigated the effects of the COVID-19 pandemic on daily behaviour and suggested five categories of factors: (1) Socio-demographic characteristics (e.g., age, educational background, occupation, marital status, nutritional status), (2) probability of COVID-19 awareness (information, seriousness, considering COVID-19 as a public health threat, probability of getting sick), (3) knowledge of COVID-19 (symptoms, prevention methods), (4) preparedness (government capability, self-preparedness), and (5) COVID-19-related behaviour (daily routine and plan change). Lins and Aquino [20] reported that panic buying behaviour might reflect psychometric properties, such as fear of being infected by COVID-19 and the fear of product scarcity. Furthermore, Alaimo and Fiore [21] stated that online shoppers with higher education levels are generally more satisfied with the online shopping experience. To assess the awareness, attitude, practice, and prevalence of COVID-19 among Riyadh residents, Alahdal et al. [22] conducted a cross-sectional survey. Several demographic factors were considered independent variables, such as gender, age, education, family size, and income. The level of awareness, attitude, and behaviour were identified and evaluated as endogenous variables. Abdullah and Dias [10] addressed the impacts of the COVID-19 pandemic and travel behaviour, collecting data from various countries worldwide. They found that people’s travel behaviour and transportation mode preferences have drastically changed due to the COVID-19 outbreak, and people’s travel intentions were mainly related to shopping. They concluded that gender, car ownership, employment status, travel distance, the primary purpose of traveling, and pandemic-related underlying factors during COVID-19 significantly influence travel behaviour during the pandemic. In the research conducted by Lehberger and Kleih [23], the reasons for the panic buying of nonperishable food during the pandemic in Germany were investigated. The “attitude”, “subjective norm”, and “fear of future unavailability” factors were found to be the main drivers of panic buying behaviour.

The review of the theoretical background of the research shows that many researchers have studied different types of human behaviour that represent societies’ economical characteristics based on the planned behaviour theory. Pre-COVID-19 or post-COVID-19, numerous studies evaluated the correlations between different types of human behaviour using the SEM method.

2.2. Research Gap

Several studies used the SEM method to assess the changes in public behaviour due to the COVID-19 pandemic. Al-Dmour and Masa’deh [24] conducted a study in Jordan on the role of social media in protecting public health against the COVID-19 pandemic. They presented “public health awareness” and “public health behavioural changes” as the mediating variables and used the SEM method to evaluate and verify them. They concluded that social media platforms might be used to enhance public protection during the pandemic. Laato et al. [25] proposed a structural model to explore how exposure to online information may lead to unusual customer behaviour during the pandemic, such as panic buying. They considered “unusual purchases” and “voluntary self-isolation” as the factors determining unusual customer behaviour and analysed the model using PLS-SEM. They suggested that consumer behaviour is well connected to the anticipated time spent in self-isolation. Moreover, they found that being overly exposed to online information sources intensifies the perceived severity of the situation and leads to unusual consumer behaviour. In a study in South Africa, Rukuni and Maziriri [26] investigated customer satisfaction with retail stores in terms of their COVID-19 readiness strategies. They implemented the SEM method using the Smart-PLS software. The results showed that the COVID-19 preparedness strategies influence customer experience of shopping from retail stores directly through the sanitisation of entrances, sanitised retail shelving, and sanitised retail counters. They concluded that customers’ satisfaction with these strategies positively affects their behavioural intentions. In another similar study, Untaru and Han [27] used SEM to explore the protective measures against COVID-19 taken by retailers and realized that these measures have a significant and positive effect on shoppers’ safety, attitude, and satisfaction, which leads to increased behavioural intentions. They also employed the metric invariance test, which found that gender, age, education, and income are moderating variables between the mentioned factors. Severo and De Guimarães [28] examined Brazil and Portugal’s socio-environmental impacts of the COVID-19 pandemic. For this purpose, they considered environmental awareness, sustainable consumption, and social actions as the variables and employed the SEM method to analyse the influence relationships among them. They claimed that the COVID-19 pandemic is a significant determinant of behavioural changes. Ghodsi et al. [29] intended to identify, analyse, and prioritise the factors influencing consumer behaviour and increasing online shopping. To achieve this, they implemented interpretive structural modelling and Microscopic–Macroscopic methods. Their analysis resulted in a five-level categorisation of the factors that influence shopping attitude and shopping trips: Level 1, age and gender; level 2, income and education; level 3, household size and COVID-19 awareness; level 4, COVID-19 attitude and COVID-19 practice; and level 5, norm subject and shopping personal control.

The literature review suggests that although several studies investigated the changes in online shopping behaviour and travel behaviour due to the pandemic separately, no study, to the best of our knowledge, has yet examined both elements in the COVID-19 pandemic situation simultaneously. Therefore, in this paper, the SEM method is used to analyse the changes in online shopping behaviour and their relationships with the changes in travel behaviour, as well as the influence of the COVID-19 pandemic on both. For this purpose, the related factors are identified, and a conceptual model of their relationships is formed using the literature review and experts’ opinions. Consequently, the SEM method is used to evaluate their reliability and validity, and eventually, the final structural model is presented and discussed.

3. Study Area

This study focuses on Tehran, the capital city of Iran. According to the Tehran population information reported by the World Capital Institute and the population study and census data from the Statisical Centre of Iran (UNFPA Iran), Tehran’s population in 2022 is around 9.4 million and has the 38th rank among the most populated cities in the world. Moreover, as reported by the Iranian Statistical Research and Training Centre in 2021, Tehran constitutes 22.2% of Iran’s total GDP and holds more than half of the country’s industrial section.

Tehran has also achieved considerable growth in e-commerce in the past few years. The charts shown in Figure 1 are collated from the 2021 report published by the Statistical Centre of Iran, the number of internet money transactions in Tehran reached more than 408 million in 2018 and with a 53% increase in 2019, reaching more than 845 million transactions. With the outbreak of COVID-19 in 2020, the number of purchase transactions grew by 40%. Considering the infrastructural developments in recent years which is indicated by the increase in the internet penetration rate, despite regressions in most industrial sectors due to the pandemic outbreak, e-commerce as a nascent segment of the economy made substantial progress.

The official announcement of the COVID-19 outbreak in Iran in late February 2020 signaled a change in the online shopping transaction trends, which shows a significant growth of 68% compared to the same period in the previous year (Figure 1). This might be mainly due to the public panic that ensued as the pandemic took place. It is worth mentioning that this drastic change occurred in March 2020, one month before the restrictions were placed on intra-city and inter-city mobility. The significance of this sudden increase in online shopping transactions may be better understood by looking at Figure 1, which shows a 12% and −10% change in the same period for March 2019 and 2021, respectively.

The traffic restrictions in Iran were applied in April 2020, which included fines on unnecessary intra-city travel after 8 pm. Only public vehicles (affiliated with municipal services, health systems, etc.) were permitted to travel without these restrictions. Moreover, inter-city travel was only allowed if authorized by the municipality of the origin or destination city. Generally, despite the society’s fragile economic conditions and the pressures on the middle-class group who represent a major portion of intra-city travel, and the financial difficulties imposed on them by the inevitable fines, these strict mobility restrictions played a key role in reducing intra-city traffic.

The city of Tehran is the main political and economic hub of Iran and almost all prominent companies and businesses have at least one branch in this region. This concentration of economic and political power in this city has resulted in the population’s high density and congestion over the past few decades. The 2020 Statistical Yearbook of Tehran reported that the city sustains 19.3 million daily intra-city trips, which usually lead to traffic bottlenecks and gridlock, and heavy rush hours in congested areas. Therefore, the development of e-commerce and e-governance infrastructures is considered a key element in intra-city transport management and travel behaviour for policymakers in a prominent metropolitan district, such as Tehran.

In densely populated cities, such as Tehran, where people spend several hours on daily trips, travel behaviour can majorly affect travel time, and less time expended on daily trips can accordingly improve quality of life. Therefore, it is important to investigate travel behaviour in these cities in the incidence of abnormal circumstances, such as the COVID-19 pandemic. Moreover, Tehran is a modern city with a considerable amount of cash flow. The people of Tehran have relatively higher incomes and consume more than other city dwellers in the country. Consequently, the influence of the pandemic on people’s shopping behaviour is more revealing in Tehran than in other cities in Iran. Therefore, the city of Tehran is a suitable sample for this study since it adequately represents large metropolitan cities around the world, and investigating the relationships between the pandemic, online shopping, and travel behaviour can produce meaningful results for policymakers.

4. Methodology

Structural equation modelling (SEM) is a label for a diverse set of methods used by scientists in both experimental and observational research across the sciences, business, and other fields [24]. As SEM comprises several techniques and analysis tools, there is not a single source for its implementation. Part of this approach originates back to the early 1900s with the development of what we now call exploratory factor analysis, usually credited to Charles Spearman (1904) [30]. After a later time, the path analysis technique was introduced to behavioural sciences and has then been broadly used, especially in sociology research [31]. In summary, the SEM group of techniques has its origins in regression analyses of observable variables (OVs) and in factor analyses of latent variables (LVs). SEM’s features include the potential to explicitly distinguish between OVs and LVs, and the capacity to analyse covariances or means in experimental or nonexperimental designs. Additionally, it brings new insight into causal inference in the behavioural sciences. This indicates that many researchers have used this model when the variables were mediators or behavioural (or psychometric) [30].

This research uses the SEM method to evaluate the correlations between the factors [32]. There are two types of variables in the SEM method. OVs are the indicators that represent the collected data, and LVs are the factors that OVs explain. A key feature that separates SEM from other statistical analysis techniques is that it can analyse both observable and latent variables. The SEM method employs three groups of tests to assess the overall fitness of the model. The first group is related to the measurement model, which is verified through the reliability and validity test. The second evaluates the structural model by calculating the paths between the LVs. The third group estimates the overall fitness of the general model [30]. The steps required for applying the SEM method in this study are described in this section and a flowchart of the SEM approach is illustrated in Figure 2.

Step 1: Defining the conceptual model and the hypotheses.

The conceptual model is comprised of the presumed correlations between the LVs. Hypothetical constructs in the conceptual model represent the relationships between the LVs in SEM. The conceptual model is then considered the basis of relationship analysis. The LVs are defined in

Table 1. Definition of the factors considered in this study.

Factor	Definition	Reference(s)
Socio-Demographic Status	The individual’s age, education level, income, and amount of daily internet use.	[29]
COVID-19 Awareness	The individual’s awareness and general knowledge of COVID-19, how it is transmitted, how it can be prevented, its symptoms, etc.	[14,22,29]
COVID-19 Attitude	The individual’s positive or negative evaluation on behaving in compliance with COVID-19-related restrictions, regulations, and instructions.	[14,22,29]
COVID-19 Practice	The level to which the individual complies with COVID-19-related restrictions, regulations, and instructions.	[14,22,29]
Subjective Norm	A social pressure perceived by the individual to do online shopping that affects their tendency to accept online shopping.	[14,29]
Online Shopping Control	The extent to which the individual feels they are in control of their online shopping. It includes the skills, capabilities, and knowledge about online shopping.	[14,18,29]
Online Shopping Attitude	The individual’s positive or negative evaluation and behavioural beliefs about online shopping.	[14,18,29]
Travel Behaviour	The individual’s daily travel habits.	[14,29]

, and the conceptual model of this study is illustrated in Figure 3.

Step 2: Data collection.

In this study, the data is collected using a 5-scale Likert questionnaire that corresponds to the OVs considered in this research. To confirm the conceptual model and to investigate the research hypotheses, an online questionnaire was employed to collect data. The sample population of this research consisted of people who had experience with online shopping and lived in Tehran, Iran for at least 5 years. To collect the data, a web link to the questionnaire was sent to potential respondents during the period between 10 May and 10 June 2022. The questionnaire was prepared in the Persian language and distributed among 330 respondents. Then, the filled-in questionnaires were reviewed to find and remove the ones whose data have a standard deviation less than 0.3 to guarantee face validity. Eventually, 274 questionnaires were deemed valid. The typical sample size in studies where SEM is used is about 200 cases [30]. The summary of the sample population’s profile and their related measurement items are presented in

Table 2. The sample profile.

Characteristic	Value	Percentage
Gender
Male	148	54%
Female	126	46%
Age (years)
<18	20	7%
18–25	36	13%
25–35	138	50%
35–45	59	22%
>45	21	8%
Education
High school or lower	27	10%
Diploma	43	16%
Bachelor’s degree	116	42%
Master’s degree	67	24%
Ph.D. degree or higher	21	8%
Household income level
Very low	22	8%
Low	40	15%
Medium	107	39%
High	78	28%
Very high	27	10%

. The table shows that 148 of the 274 respondents in this study (56%) were male, and 197 (72%) were aged between 25 and 45 years old. Additionally, 183 of the 274 respondents (67%) have a Bachelor’s or Master’s degree. Moreover, according to the collected information, people with a medium income level constitute the largest proportion in this study. The questions are presented in

Table 3. The questions and their factor loading value.

Latent Variable	Index	Observable Variable	Factor Loading
Socio-Demographic Status	SoDe-1	What is your age?	0.809
	SoDe-2	What is the highest degree or level of education you have completed?	0.772
	SoDe-3	What is your annual household income?	0.787
	SoDe-4	How much time do you spend using the Internet daily?	0.776
COVID-19 Awareness	CoAw-1	COVID-19 is transmitted through coughing.	0.792
	CoAw-2	COVID-19 is transmitted through touching and handshake.	0.724
	CoAw-3	It is possible to get COVID-19 more than once.	0.786
	CoAw-4	Children do not get infected with COVID-19.	0.771
	CoAw-5	COVID-19 can be cured using antibiotics.	0.723
COVID-19 Attitude	CoAt-1	Staying home can prevent COVID-19 spread.	0.808
	CoAt-2	Quarantining the people infected with COVID-19 prevents its spread.	0.749
	CoAt-3	Restrictions on intra-city trips prevent the spread of COVID-19.	0.788
COVID-19 Practice	CoPr-1	I wash my hands to prevent being infected by COVID-19.	0.814
	CoPr-2	I keep my hands away from my eyes, nose, and mouth to prevent being infected by COVID-19.	0.767
	CoPr-3	I wear a mask in public places to prevent COVID-19 infection.	0.761
	CoPr-4	I sanitise my groceries to prevent the chance of COVID-19 infection.	0.764
Online Shopping Attitude	ShAt-1	Online shopping is time-saving.	0.768
	ShAt-2	The wide variety of products on the Internet is satisfactory compared to stores and shops.	0.792
	ShAt-3	I find it difficult to return the product I bought online if I am unsatisfied with it.	0.825
Online Shopping Control	ShCo-1	Online shopping makes it easy to compare products and select the most suitable one.	0.754
	ShCo-2	I prefer to make my purchase from a website that provides adequate, accurate, and correct information about products.	0.807
	ShCo-3	Online shopping can reduce city traffic.	0.800
Subjective Norm	SuNo-1	I can easily make my purchase from online shops.	0.797
	SuNo-2	I receive my online purchases on time.	0.783
	SuNo-3	The prices of the products that I buy online are reasonable.	0.745
Travel Behaviour	PaTr-1	How much time do you spend on intra-city trips daily?	0.767
	PaTr-2	How much time does your family spend on daily in-person shopping?	0.772
	PaTr-3	How much time in a day do you spend on non-business intra-city trips?	0.766

.

Step 3: The measurement model validity and reliability.

The measurement model investigates the relationships between each LV and its OVs [33]. In the SEM method, the reliability is first determined to specify the suitability of the questions for measuring the LV. The SEM method has several criteria to evaluate the measurement model by calculating the reliability and validity values. These criteria are described in the following.

Cronbach’s alpha coefficient: It is a conventional criterion for measuring reliability. Its value is in the range of (−1,1), and values above 0.6 indicate acceptable reliability [34].

Composite reliability (CR): Similar to Cronbach’s alpha, this criterion measures internal consistency. It is used to check whether the OVs depend on a single factor. CR ranges between 0 and 1, and CR values above 0.7 are acceptable [35].

Average variance extracted (AVE): The average of the squared standardized pattern coefficients for OVs that depend on the same factor but are specified to measure no other factors. AVE can range from 0 to 1, and values above 0.5 are acceptable [30]. The values for Cronbach’s alpha coefficient, CR, and AVE are calculated and presented in

Table 4. The values of Cronbach’s alpha, CR, and AVE.

Variables	Cronbach’s Alpha	Composite Reliability	(AVE)
Socio-Demographic Status	0.794	0.866	0.618
COVID-19 Awareness	0.817	0.872	0.578
COVID-19 Attitude	0.682	0.825	0.612
COVID-19 Practice	0.781	0.859	0.603
Online Shopping Control	0.694	0.830	0.620
Subjective Norm	0.668	0.819	0.601
Online Shopping Attitude	0.710	0.838	0.632
Travel Behaviour	0.653	0.812	0.590

. As shown in the table, all values are in the acceptable range.

Factor loading coefficients: Factor loadings are obtained by calculating the correlation of an LV’s questions with the LV. Values equal to or greater than 0.7 indicate that the variance between the LV and its questions is more significant than the LV’s measurement error, and the measurement model’s reliability is acceptable [30]. The factor loading values obtained for the model in this study are above the critical value of 0.7. The factor loading values are depicted in Table 3.

Forner-Larcker criteria: These criteria indicate that the square root of the AVE for each LV must be greater than the correlation values of the variable with other LVs [36]. As shown in

Table 5. The Forner-Larker matrix.

	Socio-Demographic Status	COVID-19 Awareness	COVID-19 Attitude	COVID-19 Practice	Online Shopping Control	Subjective Norms	Online Shopping Attitude	Travel Behaviour
Socio-demographic status	0.786
COVID-19 awareness	0. 763	0.761
COVID-19 attitude	0.699	0.759	0.782
COVID-19 practice	0.693	0.707	0.781	0.777
Online shopping control	0.648	0.698	0.589	0.547	0.787
Subjective norms	0.762	0.732	0.713	0.719	0.753	0.775
Online shopping attitude	0.661	0.655	0.583	0.550	0.738	0.746	0.795
Travel behaviour	0.636	0.645	0.616	0.617	0.693	0.772	0.693	0.768

, the Forner-Larcker criterion is satisfied in all structures.

Step 4: The structural model evaluation.

The structural model indicates the relationships between the LVs. The structural model evaluation uses several criteria, which are explained in the following.

The Coefficient of determination (R²): This criterion determines the level of changes in a dependent LV caused by its corresponding independent LVs. The changes in R² values indicate whether an exogenous LV (independent LV) significantly impacts an endogenous LV (dependent LV). The value of R² can be in the range of (0,1), and higher values are more desirable. R² values above 0.67 are desirable, values around 0.33 are mediocre, and values around 0.190 are weak [35].

In this study, the obtained R² values for all dependent variables are in the mediocre or desirable zones. As shown in

Table 6. The matrix for the coefficient of determination.

Endogenous Variables	R2 Value
COVID-19 Awareness	0.583
COVID-19 Attitude	0.611
COVID-19 Practice	0.652
Online Shopping Control	0.422
Subjective Norm	0.678
Online Shopping Attitude	0.640
Travel Behaviour	0.638

, the R² results suggest that the conceptual model’s variables are defined correctly, and these variables introduce a high level of changes in endogenous latent variables.

Predictive relevance (Q²): Models with acceptable structural fitness must be capable of predicting the criteria related to the endogenous latent variables. This capability is evaluated by the nonparametric Stone-Geisser test, which is implemented through the Blindfolding feature in the Smart-PLS software [35]. The structural model lacks predictive relevance if the Q² value is less than or equal to zero. For Q² values of 0.02, 0.15, and 0.35, the model’s predictive relevance is small, medium, and large, respectively [37].

Concerning the model’s predictive performance, Q² values for the 0.15 and 0.35 values are medium and weak, respectively.

Table 7. The values for the Q² criterion.

Endogenous Variables	Q2 Value
COVID-19 Awareness	0.332
COVID-19 Attitude	0.366
COVID-19 Practice	0.389
Online Shopping Control	0.253
Subjective Norm	0.399
Online Shopping Attitude	0.385
Travel Behaviour	0.370

shows the Q² values for each dependent variable.

The Q² values obtained for the endogenous variables in this study are all in the desirable range. This indicates that the conceptual model’s predictive performance is at a desirable level, which means that if this conceptual model is employed on other statistical populations, the same results are likely to be obtained. Therefore, the model has acceptable credibility.

The two-tailed test: The relationships between the model’s variables may be approved or disapproved by comparing the t-value for each defined relationship. In this regard, the two-tailed test is conducted at the 0.05 level (p < 0.05), and if the absolute value of the t-value is more significant than 1.96, then the relationship between the two variables is significant. The Smart-PLS software applies this statistic through the Bootstrapping feature [37].

Figure 4 illustrates the results of the t-value significance test on the conceptual model in this study. The values on the arrows indicate the t-value for the corresponding relationship. As seen in Figure 4, the majority of relationships are at the significance level.

Step 5: The global fitness validation.

The general model includes both the measurement model and the structural model. The Goodness of Fit (GoF) criterion is used to evaluate the fitness of the general model. The GoF criterion was first introduced by [37] and is calculated as shown in Equation (1):

$$GoF=\sqrt{\overline{AVE}\ast \overline{R^2}}$$

(1)

In general, $\overline{AVE}$ indicates the average of the LVs’ AVEs and $\overline{R^2}$ is the average of R² values of the model’s LVs, as shown in the circles in Figure 4.

The values 0.01, 0.25, and 0.36 are considered the weak, medium, and strong values for GoF. This indicates that if a model’s GoF is around 0.01, then the total fitness of the model is weak, and the relationships between the structures must be modified. The model’s goodness of fit is acceptable if the GoF value is greater than 0.36 [38]. The GoF value for the model in this study is 0.606, which indicates the excellent fitness of the model.

Based on the data analysis algorithm in the Smart-PLS software, after analysing the fitness of the measurement, structural, and general models, the research hypotheses are investigated. For this purpose, the two criteria of t-value and path coefficient are used. Generally, after verifying the significance of a relationship between two variables using the t-value, the path coefficient is calculated to determine the extent of the relationship.

Figure 4 shows the final model in which bold values indicate t-values, italic values indicate path coefficients, and the values in circles indicate R². According to the values obtained for the t-value, the majority of defined relationships are found to be significant. As seen in the figure, the t-value of the approved hypotheses is more significant than 1.96.

Table 8. Summary of the model’s test result.

Hypothesis	Relationship			Path Coefficient	t Value	Test Result
1	Socio-Demographic Status	→	Covid Awareness	0.764	27.366	Approved
2	Socio-Demographic Status	→	COVID-19 Attitude	0.283	3.827	Approved
3	Socio-Demographic Status	→	COVID-19 Practice	0.289	6.348	Approved
4	Socio-Demographic Status	→	Online Shopping Control	0.65	18.952	Approved
5	Socio-Demographic Status	→	Subjective Norms	0.371	6.652	Approved
6	Socio-Demographic Status	→	Online Shopping Attitude	0.134	2.281	Approved
7	COVID-19 Awareness	→	COVID-19 Attitude	0.544	7.775	Approved
8	COVID-19 Awareness	→	Subjective Norms	0.187	2.812	Approved
9	COVID-19 Awareness	→	Online Shopping Attitude	0.072	0.974	Disapproved
10	COVID-19 Attitude	→	COVID-19 Practice	0.579	12.691	Approved
11	COVID-19 Attitude	→	Subjective Norms	0.139	2.001	Approved
12	COVID-19 Attitude	→	Online Shopping Attitude	0.018	0.287	Disapproved
13	COVID-19 Practice	→	Subjective Norms	0.22	2.73	Approved
14	COVID-19 Practice	→	Online Shopping Attitude	−0.053	0.944	Disapproved
15	Online Shopping Control	→	Online Shopping Attitude	0.357	6.236	Approved
16	Online Shopping Control	→	Travel Behaviour	0.179	2.353	Approved
17	Subjective Norms	→	Online Shopping Attitude	0.346	5.344	Approved
18	Subjective Norms	→	Travel Behaviour	0.493	7.498	Approved
19	Shopping Attitude	→	Travel Behaviour	0.193	2.971	Approved

depicts a summary of the conceptual model’s analysis, in which the result of analysing each hypothesis is presented.

5. Discussion

The final structural equation model, shown in Figure 4, suggests that the factors considered in this study may be signified in four stages, and each stage may be defined by the variable(s) it is composed of. The first stage includes the socio-demographic status, which indicates the characteristics of the population (age, education level, income, etc.), and as it is situated at the highest level of independence, it has a profound effect on other variables of the model. The socio-demographic factor is broadly used in the literature in similar studies as an independent variable. The second stage comprises COVID-19 variables, which denote the pandemic conditions imposed on people’s daily activities and affect their suppositions and behaviour. There are three variables, including, in order of their effect on each other, COVID-19 awareness, COVID-19 attitude, and COVID-19 practice. This order of dependency among these three variables has also been discovered by Ghodsi et al. [29] and Alahdal et al. [22]. As seen in Figure 4, this stage is greatly affected by the first stage, and consequently, it affects the elements in the third stage related to online shopping behaviour. The third stage includes subjective norms, online shopping control, and shopping attitude. This arrangement of variables in stage three constitutes the “online shopping behaviour” factor, also confirmed by Ajzen [14]. Moreover, other than being affected by the second stage (COVID-19 conditions), stage three is also affected by the first stage (socio-demographic status), and it affects the fourth stage (travel behaviour), which is the target variable of this study.

By defining these four stages and indicating their relationships, it is now possible to observe how the variables in each stage affect each other. Figure 4 shows a high coefficient value of socio-demographic status and COVID-19 awareness, indicating a strong relationship between these two variables. COVID-19 awareness suggests an individual’s conception of the potential transmission modes of COVID-19, its symptoms, groups of people more vulnerable to the virus, and its mortality rate. It involves an inner aspect, personal knowledge, and an outer aspect, social knowledge (the mainstream media, scientific sources, etc.). Nonetheless, the effectiveness of propagating health advice through mass media to increase public awareness is closely tied to socio-demographic status; namely, public awareness of COVID-19 depends mainly on people’s social and economic positions. Bearing this in mind, it seems reasonable to anticipate that people with higher awareness of the pandemic conditions display higher tendencies and take more responsibility to alleviate its spread. This anticipation is confirmed by the significance of the relationship between COVID-19 awareness and COVID-19 attitude. The next significant relationship in this stage is between COVID-19 attitude and COVID-19 practice. It can be inferred from this relationship that having a strong attitude toward the pandemic conditions eventually leads people to better comply with the regulations that aim to reduce the virus’s prevalence. In addition to the relationships between the variables in the second stage, their relationships with the variables in stage three are investigated. The final structural model shows that the impacts of the COVID-19 variables (second stage) on the online shopping attitude are insignificant. Nevertheless, the COVID-19 variables indirectly affect the online shopping attitude through the relationship they have with subjective norms; namely, the pandemic conditions majorly affect subjective norms.

The variables in stage three represent online shopping behaviour. As shown in Figure 4, they are affected by the variables in the first and second stages and affect the variables in the fourth stage. The strong relationship between the socio-demographic status and online shopping behaviour variables indicates the significant effect of people’s social and economic position on online shopping, amplified by pandemic conditions. Furthermore, the COVID-19 variable, especially COVID-19 awareness, plays a prominent role in the formation of subjective norms, and they consequently affect online shopping behaviour. This also justifies the impact of subjective norms and online shopping control on the online shopping attitude. It is worth mentioning that although the online shopping attitude does not have a significant relationship with the COVID-19 variables, it is indirectly affected by them. The final model suggests that by the outbreak of the pandemic, the more people are exposed to the perceived social pressure of online shopping (subjective norms), and the more confident they are with their ability to shop online (online shopping control), the more of a positive attitude they have toward online shopping (online shopping attitude).

Eventually, regarding the target variable of this study, the significant relationship between the online shopping variable and travel behaviour (the fourth stage) indicates that by improving the variables in the third stage, travel behaviour is also enhanced. This indicates that higher perceived pressure from society to shop online, stronger propensities for online shopping, and better capabilities to shop online will lower the frequency of intra-city travel. In a broader view, the more people use online shopping platforms to satisfy their shopping needs during the pandemic, the less they are inclined to make intra-city trips.

The SEM approach, used in this study to determine the relationships between COVID-19, online shopping, and travel behaviour variables, was implemented in the city of Tehran, Iran. As the capital city of Iran, Tehran shoulders a large fraction of Iran’s economy and is the largest business hub of the country. Therefore, it properly exemplifies the world’s populous cities in which the citizen’s travel behaviour underlies the city’s economic ecosystem. It was found in this study that COVID-19 awareness directly and substantially influences consumers’ online shopping behaviour and indirectly influences travel behaviour. This result aligns with what really happened at the advent of the pandemic in Tehran. In March 2020, the country saw a bulge in internet money transactions. This sudden surge in online purchase transactions was not triggered by the travel restrictions since the intra-city mobility restrictions were placed on April 2020. Therefore, this increase in transactions justifies the panic breaking out due to the pandemic. The COVID-19 conditions drastically influenced the citizens’ online shopping behaviour resulting in an increase in the number of online monetary transactions. It was also revealed that the prominent socio-demographic status is the mid-aged, mid-income, and mid-education categories. Therefore, to decrease the number of intra-city daily trips, policymakers should target their COVID-19 awareness-raising policies on adults with medium income and education. However, they should simultaneously facilitate the usage of online shopping platforms in order that people are less inclined to shop in person. The results of the study suggest that online shopping control and online shopping attitude influence travel behaviour both directly and indirectly. Moreover, the data regarding the internet penetration rate in Tehran (Figure 1) show an increase in the past few years. Policymakers can take advantage of the improvement in information and communication structures and promote investment in online shopping platforms to increase people’s inclination to online shopping and persuade them to improve their skills, capabilities, and knowledge about online shopping. This will consequently result in fewer daily trips.

6. Conclusions

First declared by the World Health Organization on 11 March 2020, the COVID-19 pandemic has fundamentally impacted human interactions, from international political and economic affairs to daily chores and routines. Shortly after the announcement of the pandemic, strict restrictions were imposed on outdoor and public activities aiming to limit face-to-face human contact to the greatest extent possible. For many, this new and unprecedented situation has led to changes in their travel behaviour to limit their daily trips to only primary purposes, one of which is shopping. Although online platforms have been significantly effective in keeping people indoors for study, work, and entertainment, the influence of online shopping during the pandemic on daily trips entails extensive and in-depth attention. For this purpose, the present study considered four stages of factors: Socio-demographic status, COVID-19-related variables, online shopping variables, and travel behaviour. The aim was first to discover the correlations between these variables and second, to find how travel behaviour is affected by other variables. For this purpose, the variables defining each stage were identified using the literature review. Then, a conceptual model of their relationships with each other was devised using experts’ opinions and the results of previous studies. These relationships were then analysed and verified using the SEM method. The validity and reliability of all but three relationships were confirmed in the final model.

A glance through the final model (Figure 4) confirms a cause-and-effect relationship between the stages. This indicates that socio-demographic status as an exogenous variable in the first stage directly or indirectly impacts other variables in other stages. The second stage pertains to the COVID-19 variables that influence the third and fourth stages. The online shopping variables in the third stage are influenced by both previous stages and directly impact the fourth stage, which is travel behaviour. In agreement with this study, Ghodsi et al. [29] also argued that demographic variables (e.g., age, gender, education, income) affect all other COVID-19 and online shopping variables, and the travel behaviour variable is last in the cause-and-effect hierarchy.

The findings of this study revealed that people’s intra-city travel behaviour is affected by their position on online shopping. Moreover, it was found that the more people are aware of the severity of the pandemic and the better they comply with COVID-19 regulations and restrictions, the more inclined they are to do online shopping. Therefore, policymakers can focus on facilitating the accessibility and usage of online shopping platforms to reduce intra-city traffic and prevent the spread of the virus. Furthermore, it might be a good strategy for the private sector, especially online retailers, to be more active in raising the public’s awareness of the COVID-19 pandemic by launching campaigns, providing free COVID-19 protective products (e.g., masks and sanitisers), and offering delivery discounts during the pandemic, to persuade consumers to buy their necessities online.

This study can be extended in several ways. The presented structural model can be used to analyse other consumer behaviour. Moreover, the factors used in this study may be modified to inspect other plausible relationships. For example, the “subjective norms” factor considered in this study is related to social norms that only lead to a tendency toward online shopping. However, social norms can also affect inclinations to comply with COVID-19-related regulations and instructions. Furthermore, other informative modelling tools may be used to study correlations between the factors. Finally, the results of this study are limited by the geographical and cultural attributes of the sample population and the regulations applied to the studied geographical zone.