Role of country image, subjective knowledge, and destination trust on travel attitude and intention during a pandemic

1. Introduction

When the SARS-CoV-2 virus started spreading around the world in late 2019 and early 2020, the responses adopted by countries varied greatly. At one extreme were countries such as China, Australia and New Zealand with a zero-Covid tolerance, locking down millions of people to prevent the spread. At the other extreme were tourism-dependent countries like Turkey and Mexico that remained open for tourism and did not impose any restrictions on gatherings until October 2020 in Turkey, while Mexico only suspended some non-essential activities. In this spectrum, the United States is positioned closer to these latter countries, especially in Republican-led states such as Florida, while Canada leaned more towards a stringent and cautious approach to handling the pandemic.

Countries' different responses to the initial spread of the coronavirus has influenced people's trust in the destination (DT) and therefore impacted visit intention (VI). The tourism industry is one of the hardest hit sectors by the COVID-19 pandemic and is eager to recover to pre-pandemic levels of activity at a minimum as the pandemic is becoming more endemic. While various factors contribute to uncertainty on what its future might hold, this study aims to explore how the image of countries (CI) along with the subjective knowledge (SK) of the pandemic that individuals believe they have builds trust toward each country, which then transfers to their overall attitude and intention to travel.

The USA and Canada had significantly different approaches in terms of controlling the spread of the virus. Thus, these two countries can be seen as representative of a broader range of countries and serve as examples of how potential travellers responded to measures taken. Both countries shut their borders in April 2020 with every jurisdiction imposing measures of varying stringency. Indeed, for the first time since Canadian Confederation in 1867, Canada and the U.S. closed the joint border to most travellers as they fought the spread of the COVID-19 virus. Canada's experience with handling the SARS coronavirus outbreak in 2003–2004, its single-payer national health insurance plan and greater control over its hospital system (Hess & Bitterman, 2020), allowed Canadian health authorities to slow the initial spread of coronavirus.

In the USA, the pandemic response quickly became political with some states imposing few, if any, restrictions. As a result, domestic tourism in the world's second largest internal market for travel continued to operate at about 45% of seat capacity (Grant, 2020). This contrasted with Canada where measures were somewhat more uniformly enforced and far more stringent, even though the number of deaths per million was much lower at 247 compared to 558 in the USA as of September 2020 (Joppe & Foti, 2020). During the first two waves of the pandemic, Canada's highest-reporting regions were low compared to places south of the border. Despite this, Canadians have been much more cautious when it comes to travel, and even a year later air traffic hovered around 5% of 2019 vol (Statistics Canada, 2021). This stark contrast raises questions about how residents in each country perceive the impact of the handling of the pandemic on their own and neighbouring country's image, their trust in the destination and intention to visit in the near and distant future.

The importance of CI in shaping consumer decision-making has long been affirmed but has never been tested under pandemic conditions. CI has also been found to be a critical factor that influences people's attitudes toward travelling within or to a country (Elliot et al., 2011). Tourism studies traditionally have considered the influence of tourism destination image (TDI) in terms of tourists' attitude and future VI (e.g., Hunt, 1975) rather than the image projected at the country level. Since the breakout of the global pandemic, a specific destination's image might not play as significant a role in the tourists' travel decision as previously. Overall perception of a country might be more critical for tourists' decisions to travel as each national government, rather than individual destinations, has imposed its own policies and restrictions on travelling across borders. Individual existing country image might therefore have been modified by observing the way each government handled the situation. In addition to the individual tourist's overall perception of a country, how much they understand the COVID-19 pandemic and its risks, which can be measured by SK, might be another major evaluating factor (Gursoy & McCleary, 2004). Research has shown that the greater the SK about a destination, the better decisions will be made, leading to an increased intention to travel (Artigas et al., 2017). However, the reverse can also be true: an increase in SK may inhibit their willingness to choose a destination (Park & Jang, 2013). Thus, an overload of information can lead individual tourists to self-assess which or how many cues are important in their decision-making (Tassiello & Tillotson, 2020). Based on both CI and SK, tourists might build (or not) their trust, which might be the most important factor considered in travel decisions during the pandemic. Unless there is a high level of DT, it is less likely anyone will travel. DT represents the emotional component of the destination (Roodurmun & Juwaheer, 2010), and building trust towards a destination is a vital stage of building VI (Al-Ansi & Han, 2019).

Although the USA and Canada have similar cultural backgrounds with economies that are interdependent to a very high degree, the differences in approach to the pandemic lend themselves to investigating whether the three factors of CI, SK, and DT, impact people's travel attitude and VI differently for domestic and international travel. By applying these well-established marketing constructs and understanding how closely they affect of each other will allow tourism organizations and stakeholders at country levels to reevaluate the way they project their images not only to their own domestic residents but also for potential travelers from other countries. To this end, their handling of the pandemic is particularly important. To the authors' best knowledge, the three key factors (CI, SK, and DT) have never been studied simultaneously for their impacts on travelers' attitude and behavioural intentions towards travelling domestically and internationally. Nor have specific tourism destinations been studied under the adverse conditions of an ongoing pandemic that is impacting all countries to different degrees. Thus, the study findings add new empirical evidence to validate the roles of these three factors in forming attitude and future behavioural intentions in a tourism context.

Thus, the objectives of this study are 1) to understand the important factors influencing domestic and international travel under conditions of an ongoing pandemic. 2) to investigate the relationships among country image, subjective knowledge of the COVID-19 pandemic, destination trust, attitude toward travelling, and travel intentions; and 3) to conduct a multi-group analysis of the relationships among the constructs between domestic and international travel intentions in the short and long term.

2. Literature review

2.1. Country image (CI)

Country image is a critical construct that has been widely discussed in previous literature. Country-of-origin or product country image has been studied from various perspectives in the international marketing literature since the 1960s (Dichter, 1962; Roth & Diamantopoulos, 2009) and its important role in consumer's evaluation of foreign products has been clearly demonstrated (Han, 1990). The concept of CI was first discussed in Nagashma's (1970) research on customer attitudes about a foreign product. The representative products of a country and CI influence the total product image. For instance, ‘Made in Japan’ denotes the best electrical appliances, while ‘Made in Germany’ stands for practicality, masculinity, and conservatism. CI is defined as ‘the total of all descriptive, inferential, and informational beliefs one has about a particular country’ (Martin & Eroglu, 1993, p. 193). It is associated with people's perception of multifaceted elements of a country, including history, economy, traditions, technology, politics, culture, business, and society (Chaulagain et al., 2019; Roth & Diamantopoulos, 2009). It is also interpreted as ‘mental maps or knowledge structures related to countries’ (Nadeau et al., 2008, p. 87). This interpretation highlights that CI is a stereotype about a nation that will influence people's overall evaluation of the country and further impact their attitudes towards its specific products (Palau-Saumell et al., 2016).

CI is conceptualized as a multi-dimensional construct, including two types of evaluation: cognitive measures are related to beliefs and affective measures are about emotional responses (Martínez & Alvarez, 2010; Yu & Zhang, 2020). The former refers to people's beliefs and opinions about the characteristics and attributes of a country (Pike & Ryan, 2004), while the latter stands for their emotions and feeling about the country (Martínez & Alvarez, 2010). Some researchers examined these components of CI (Buhmann & Ingenhoff, 2015; Yu & Zhang, 2020) and have highlighted the significant importance of cognitive component relative to the affective one (Fakeye & Crompton, 1991; Phillips et al., 2013). Martin and Eroglu (1993) developed a 14-item scale to measure cognitive CI, frequently cited in tourism studies. Stepchenkova and Shichkova (2017) put forward a framework to measure people's evaluation of different elements (e.g., climate, people, history, crime) of a country. Woo et al. (2017) examined CI based on the components of overall image, people, products, desired intention. Zhang et al. (2016) invited respondents to evaluate China's CI using a 13-item scale, which is also a cognitive measure. Overall, most of the scales measuring CI lack an affective component (Roth & Diamantopoulos, 2009; Martínez & Alvarez, 2010).

Destination, as one of the products that ‘originate’ in a country, also shares some image components with the general CI (Stepchenkova & Shichkova, 2017). Although CI has not been investigated as intensively as TDI, both are deemed dynamic concepts and are based on the evaluation of attributes, are multi-dimensional and constitute a subjective assessment (Martínez & Alvarez, 2010). Compared to TDI, CI plays a more important role in understanding the impacts the outbreak of COVID-19 had on people's perceptions and intentions (Chen et al., 2020; Simons, 2020). CI represents a combination of diverse elements that are independent of a specific context, but TDI is from the tourist's perspective and indicates the specific area visited, such as a tourist attraction, a city or a region (Mossberg & Kleppe, 2005; Nadeau et al., 2008). The main distinction between the two is that CI is an overall evaluation and has a broader view (Martínez & Alvarez, 2010; Ramkissoon & Nunkoo, 2011). How a country deals with the pandemic is based on diverse considerations, including technology, health care system, and culture, instead of the tourism service offerings at a destination. Therefore, CI is a critical concept that has been associated with adverse events like SARS and COVID-19 (Chen et al., 2020).

2.2. Subjective knowledge (SK)

Researchers have proposed three types of consumer knowledge, including usage experience, objective knowledge, and SK (Hem & Iversen, 2009). The first category refers to the amount of customer's experience of purchasing or using a product; the second to the amount of knowledge a customer actually has stored in memory; and the third measures the customer's perception of how much they know. Among the three types, SK was found to play a more important role in decision-making (Ellen, 1994; House et al., 2004; Raju et al., 1993). It is also argued that measuring SK is easier than examining objective knowledge (Flynn & Goldsmith, 1999). Objective tests of a consumer's knowledge need to be individual-based and fact-checked, while SK can be investigated by asking what people think they know about the product. Recent studies have focused on understanding the relationship between SK and consumer behaviours (Hem & Iversen, 2009; Phillips et al., 2013; Utkarsh et al., 2019). Despite SK being a critical concept in consumer behaviour, there are only a few definitions proposed by researchers (Flynn & Goldsmith, 1999). All highlight that SK is a customer's perceived level of their knowledge and self-confidence (Dodd, Laverie, Wilcox, & Duhan, 2005).

There is a direct relationship between SK and consumer behaviours such as information search (Dodd et al., 2005; Ramkissoon & Uysal, 2011), brand evaluations (Hem & Iversen, 2009), information processing (Utkarsh et al., 2019), decision processes (Moorman et al., 2004), and behavioural intention (Phillips et al., 2013). Customers with higher levels of SK rely on impersonal information (e.g., books, news) when making decisions and are less influenced by personal sources (e.g., acquaintances, friends). They are confident using impersonal sources and their existing knowledge (Dodd et al., 2005). SK can also strengthen the relationship between consumer self-confidence and intention for information search and dissemination (Utkarsh et al., 2019).

2.3. Destination trust (DT)

Tourism researchers conceptualize DT as ‘a multidimensional construct, including the local inhabitants and public and private institutions that are honest, benevolent and competent’ (Marinao et al., 2012) or the confidence and certainty perceived by tourists toward the tourism services or offerings (Al-Ansi & Han, 2019). Trust is an effective approach to minimizing the perception of uncertainty and risk (Han & Hyun, 2013; Pavlou et al., 2007). Therefore, tourists believe a trustworthy destination can provide transparency, reliability, less-risk, and less-hassle services and experiences (Roodurmun & Juwaheer, 2010).

Customer trust or brand trust has been intensively discussed in the customer behaviour and marketing area and is essential in building strong customer relationships and sustainable market share (Urban et al., 2000). Sirdeshmukh et al. (2002) developed a framework and proposed that customer trust significantly influences perceived value and then impacts customer loyalty. Trust develops customers’ emotional attachment toward a brand (Esch et al., 2006) and thus it has a positive effect on perceived corporate social responsibility and purchase intention (Nguyen & Pham, 2018). It also leads to repeat purchases as it effectively minimizes uncertainty during online shopping (Chiu et al., 2012; Kim & Oh, 2002).

In tourism, trust studies mainly examined either organizational trust and interpersonal trust (Abubakar et al., 2017). The former refers to trust in destinations, tourism companies, agencies, and government (Kim et al., 2011; Nunkoo et al., 2012; Razak et al., 2014), while the latter stands for trust between people such as servants, residents, tourists, and travel companions (Chang, 2014; Ouyang et al., 2017). For example, residents' interpersonal trust can influence place attachment and pro-environmental behavior (Ramkissoon, 2020). Based on these two categories, Liu et al. (2019) developed a scale for trust, including the dimensions of trust in scenic spot, administration, agency, employees, residents, and other tourists. Trust is especially important for tourism because of the simultaneous nature of production and consumption of services and products (Abubakar et al., 2017; Abubakar & Ilkan, 2016). Abubakar et al. (2017) constructed destination trust in the med-tour context and specified it as the tourists' trust in Turkish hospitals and healthcare services. Researchers have investigated the relationships between information, DT, VI, satisfaction, and loyalty (Artigas et al., 2017). Information is a factor affecting tourists' perceived DT, and DT significantly impacts tourists' VI (Abubakar & Ilkan, 2016). Additionally, it positively influences tourist satisfaction and loyalty towards a tourist destination (Artigas et al., 2017). Although DT is a critical construct, there is a lack of understanding how it influences tourists' travel attitude and VI during an adverse event, such as a pandemic. Therefore, this research focuses on understanding the impact of trust in the destination's handling of the pandemic, hereafter referred to as destination trust.

2.4. Hypotheses development

A sociological theory of trust proposed by Luhmann (1979) is the theoretical foundation of this research. This theory puts forward that trust consists of three modes, including familiarity, confidence, and trust. Familiarity and confidence play a critical role in building trust (Luhmann, 1979), which only exists in a familiar world. Luhmann's confidence is interpreted as “a mix of cognitive and emotional perceptions” (Elliott & Yannopoulou, 2007, p. 989). Drawn from this theory, subjective knowledge should be the base of mode of familiarity, thus, they are linked. Mode of confidence consists of cognitive and emotional perceptions as country image also includes cognitive and affective perceptions of a country, they could be linked. Previous studies have proposed the relationships between brand knowledge, brand image and brand trust based on Luhmann's (1979) theory (Hsu & Cai, 2009).

Attitude is a persistent psychological construct that can effectively influence human behavior and it has been used as a predictor for intention (Ajzen, 1991; Kraus, 1995). According to the Theory of Planned Behavior, the degree to which an individual has a favorable or unfavorable evaluation of the behavior determines their intention (Ajzen, 1991). Travel attitude has been identified as a reliable predictor to VI (Huang & Hsu, 2009). Researchers have investigated the antecedents of travel attitudes and identified information as playing a considerable role in influencing and forming attitudes (Jalilvand et al., 2012). Attitude has also been treated as an important moderator in diverse contexts. For example, there are a myriad of studies examining knowledge-attitude-behavior model in the public health field (Hu et al., 2016; Liu et al., 2016; Ramkissoon, 2021). Ramkissoon (2021) discussed that unclear information can lead to distrust in vaccines in general and negative attitudes towards COVID-19 vaccine in particular. As a result, people will be less likely to receive a vaccine. It shows the critical influence of knowledge on attitude and behavior. In the tourism context, the destination image-attitude-intention model is a classic framework (Jalilvand et al., 2012; Kim & Kwon, 2018). Researchers have investigated how destination image can significantly impact travel attitude and intention. Additionally, destination image significantly influences place attachment and impacts perceptions of experiences (Jiang et al., 2017). Researchers in the digital commerce and hospitality areas have proposed the trust-attitude-intention model, where attitude is a critical mediator in the relationship between trust and intention (Palacios-Florencio et al., 2018; Wu & Liu, 2007). Therefore, this research included attitude as an important mediator and structured the framework based on previous studies with knowledge, destination image, and trust as three antecedents of travel attitude.

A country-of-origin (COO) and a country stereotyping effect make up a country's widely held image. The COO effect relates to how products manufactured in that country are perceived, and in the tourism context, it explains how people's perception of a country affects their attitudes towards it as a destination (Elliot et al., 2011; Lee & Lockshin, 2012). Country stereotyping influences decision-making when consumers are not familiar with the product. Hence, customers perceived CI determines their perception of the quality of the products from that country (Chung & Chen, 2018). The impact of CI on customer behavior has also been discussed in the tourism literature (Chaulagain et al., 2019; Chung & Chen, 2018). CI is perceived as a stereotype and has a halo effect; that is, people's overall evaluation of the country will directly influence their attitudes towards its products and intention to use (Palau-Saumell et al., 2016). Thus, a favorable CI enhances tourists' intention to visit the destination (Chaulagain et al., 2019). The affective and cognitive components of CI significantly affect destination beliefs, and further impact destination receptivity (Elliot et al., 2011). DT is a type of destination belief, which will be influenced by CI. Therefore, the following hypotheses were proposed:

H1

Country image (CI) positively influences travel attitude (TA).

H2

Country image (CI) positively influences destination trust (DT).

H3

Country image (CI) positively influences visit intention (VI).

SK is about customers’ perceived level of their knowledge, with impersonal information being one of the main factors that influence it (Dodd et al., 2005). The severity of the pandemic has been broadcasted through multiple information channels (Paraschi, 2020). For example, both the Canadian and American governments have published travel advisories to suggest residents avoid non-essential travel (Government of Destination Canada, 2021; U.S. Department of State, 2021). People with higher levels of knowledge about a risk will be more cautious about travelling (Sharifpour et al., 2014) as it will allow them to evaluate the attributes, benefits, and uncertainties of a destination (Ratchford, 2001). If the destination is seen to have higher risks or is unable to control the pandemic effectively, potential tourists will not trust it, exhibiting negative attitudes (Bae & Chang, 2020), and not intent to travel. The effects of SK on intention have also been evidenced by previous studies (Phillips et al., 2013). Thus, leading to these hypotheses:

H4

Subjective knowledge (SK) of COVID-19 negatively influences destination trust (DT).

H5

Subjective knowledge (SK) of COVID-19 negatively influences travel attitude (TA).

H6

Subjective knowledge (SK) of COVID-19 negatively influences visit intention (VI).

DT reduces perceived uncertainty and risks in a destination where customers feel vulnerable (Liu et al., 2019). Tourism, especially international tourism, is considered to be particularly sensitive to global risk factors (Ritchie, 2004), sensitivity that is heightened during a crisis (Pizam & Mansfeld, 1996) such as COVID-19. Previous research confirms that DT influences tourists’ risk perceptions (Kim et al., 2009) and emotional attachment to a destination (Chen & Phou, 2013). In tourism context, tourists are more likely to have positive attitude and visit destinations that they deem dependable and trustworthy (Ekinci & Hosany, 2006; Han & Hyun, 2013; Roodurmun & Juwaheer, 2010). Tourists may even establish emotional attachment to a trustworthy destination in the long term (Thomson et al., 2005), leading to the following hypotheses:

H7

Destination trust (DT) positively influences travel attitude (TA).

H8

Destination trust (DT) positively influences visit intention (VI).

The relationship between travel attitude and visit intention has been intensively discussed in previous literature. According to Ajzen's (1985) theory of planned behavior, positive attitude leads to positive behaviour. Therefore, tourists with positive travel attitudes towards a destination will be more likely to visit that destination (Huang & van der Veen, 2019). A large number of empirical studies confirm that travel attitude is a strong predictor of visit intention (Lam & Hsu, 2006; Letheren et al., 2017). Hence, proposing the last hypothesis:

H9

Travel attitude (TA) positively influences visit intention (VI).

3. Methodology

3.1. Survey design

This research adopted scales used in previous literature. The survey included two main parts: the first contained questions about CI, DT, SK, travel attitude, and visit intention, while the second asked about participants' demographic information. To measure CI, 18 items were adopted from previous studies (Chaulagain et al., 2019; Martin & Eroglu, 1993; Martínez & Alvarez, 2010). The 6-item scale of DT was adopted and adapted from Sirdeshmukh et al. (2002) and Nguyen and Pham (2018). DT in this study specifically refers to trust in the destination's handling of the pandemic. SK (Flynn & Goldsmith, 1999) and travel attitude (Huang & van der Veen, 2019) were measured using five items respectively. Two timeframes (i.e., one year and two years) were used to better understand participants' short-term and long-term visit intention. Each timeframe was examined through three intention items (Lee & Lockshin, 2012). All measurement items used a 7-point scale. Four sets of questionnaires were designed to investigate residents' perceptions and visit intention regarding domestic and international travel during COVID-19: The first examined Canadian domestic travel; the second American domestic travel; the third examined Canadian travel to the US; and the fourth American travel to Canada. The wording of the scales were adjusted based on the target group and context.

3.2. Data collection and analysis

To reach a wide range of participants, data were collected through the online panel Dynata. The target population was Canadian and American adults 18 years and older, who were residents of either country and who had traveled in the last two years. In addition, the international surveys asked about having visited Canada/the US in the last 10 years. A pilot study was conducted to identify ambiguous or misleading questions. Data were formally collected in early August 2020 using stratified sampling of 500 respondents per questionnaire equally divided by gender. In addition, the questionnaires were distributed according to the geo-locations of respondents (i.e., Provinces/Territories in Canada; States in the USA) to collect the perceptions from a broader group of people and to enhance the generalizability of this research. The distribution quota was based on the national population census in Dynata's database. SPSS 26 and AMOS data analysis software were used to conduct exploratory factor analysis (EFA), confirmatory factor analysis (CFA), structural equation modeling (SEM), and multigroup analysis.

4. Results

4.1. Respondent profile and attitudes towards the handling of the pandemic

As the stratified sampling method was adopted, there was an equal number of male and female respondents. Most respondents were between 25 and 64 years old: This age group for Canadians in Canada, Americans in the US, Canadians to the US, and Americans to Canada accounted for 70.6%, 70.4%, 73.6%, and 65.2%, respectively. The largest group of respondents had a college or university degree. Around 30%–40% of respondents had an average household income of over USD $100,000. More than half of the respondents were married or living with a partner. Details can be found in the supplemental material online.

Additionally, respondents' attitudes towards how their own country's handling of the pandemic were examined. For the statement ‘Canada/US is competent at handling the COVID-19 crisis’, respondents indicated their agreement using a 7-point scale (1 = strongly disagree, 7 = strongly agree). Respondents believed that Canada was much more competent at handling the COVID-19 crisis: Mean _Canada = 5.38, S.D. _Canada = 1.26; Mean _USA = 3.10, S.D. _USA = 2.05 (F = 892.94, p < 0.001).

4.2. Exploratory factor analysis (EFA)

To explore the factorial structure of CI, 18 items of the construct were analyzed using exploratory factor analysis with varimax rotation, details of which can be found in the supplemental material. The Kaiser-Meyer-Olkin results suggested that the items are adequate for factor analysis: KMO = 0.945, Bartlett's test of sphericity $x^2$ (105) = 28709.686, p < 0.001. The maximum likelihood factor analysis with a cut-off point of .60 and the Kaiser's criterion of eigenvalues greater than 1 yielded a three-factor solution as the best fit for the data (Stevens, 1992). Three items were excluded because of low factor loadings: an economically stable country (0.591), a socially developed country (0.557), and a politically stable country (0.582). The remaining 15 items were loaded in three factors explaining 77.2% of the total variance.

4.3. Confirmatory factor analysis (CFA)

The CFA was conducted to examine the structure of each latent variable. Considering the significant differences, two separate CFA was done for the domestic travel and international travel datasets.

4.3.1. Domestic travel

CI was treated as a second-order construct, the composite reliability, convergent validity, and discriminant validity were first examined. Factor loadings greater than 0.70 were remained (Hair et al., 2010). Three items were excluded due to low factor loadings: an important country (0.644), a politically influential country (0.635), and a country that respects the environment (0.677). The standardized coefficients of the rest of the 12 items ranged from 0.714 to 0.921. The three latent variables of CI had satisfactory composite reliabilities (CR), as all values were above the cut-off point of 0.7 (see Table 1 ). The average variance extracted values (AVEs) of F1, F2, and F3 were all greater than 0.5, showing good convergent validity. Additionally, the square root of a construct's AVE was larger than its correlation with other constructs: $\sqrt{{\text{AVE}}_{\text{F}1}}=0.852$, $\sqrt{{\text{AVE}}_{\text{F}2}}=0.862$, $\sqrt{{\text{AVE}}_{\text{F}3}}=0.773$. Therefore, discriminant validity was achieved (Anderson & Gerbing, 1988). The model fit indices showed a good fit: ${\chi }^2$ = 328.358 (df = 43), ${\chi }^2$/df = 7.636, p < 0.001; goodness of fit index (GFI) = 0.948, incremental fit index (IFI) = 0.973, normed fit index (NFI) = 0.969, comparative fit index (CFI) = 0.973, and root mean square error of approximation (RMSEA) = 0.082, all exceeding required cut-off value (Anderson & Gerbing, 1988).

Table 1.

Correlation matrix of country image factors.

Constructs	F1	F2	F3	Mean	SD	CR	AVE
F1	1			5.443	0.058	0.930	0.726
F2	0.851	1		5.563	0.118	0.920	0.743
F3	0.680	0.707	1	5.763	0.136	0.815	0.597

Note: S.D. = Standard deviation; C.R. = Composite reliability; AVE = Average variance extracted.

After the second-order construct was confirmed, the composite reliability, convergent validity, discriminant validity, and model fit of the latent constructs including CI, SK, DT, travel attitude (TA), and visit intention within one year (VI1) were tested. The CRs and AVEs of the five latent constructs were above the cut-off point of 0.7 and 0.5, respectively (see Table 2 ). Additionally, the square root of a construct's AVE was larger than its correlation with other constructs: $\sqrt{{\text{AVE}}_{\text{CI}}}=0.876$, $\sqrt{{\text{AVE}}_{\text{DT}}}=0.935$, $\sqrt{{\text{AVE}}_{\text{SK}}}=0.854$, $\sqrt{{\text{AVE}}_{\text{TA}}}=0.946,\text{ and }\sqrt{{\text{AVE}}_{\text{VI}1}}=0.914$. Therefore, the constructs had both good convergent and discriminant validity. This model also had a good model fit: ${\chi }^2$ = 1570.973 (df = 354), ${\chi }^2$/df = 4.438, p < 0.001; GFI = 0.900, IFI = 0.964, NFI = 0.955, and CFI = 0.964; RMSEA = 0.059.

Table 2.

Correlation matrix of the five constructs.

Constructs	CI	SK	DT	TA	VI1	Mean	SD	CR	AVE
CI	1					5.563	1.121	0.908	0.767
SK	−0.011	1				4.701	1.504	0.890	0.730
DT	0.712	−0.211	1			4.604	1.790	0.977	0.874
TA	0.514	−0.254	0.605	1		4.427	1.623	0.977	0.895
VI1	0.218	−0.056	0.166	0.364	1	4.814	1.722	0.938	0.835

Note: CI = country image; SK = subjective knowledge; DT = destination trust; TA = tourist attitude; VI1 = visit intention within one year; S.D. = Standard deviation; C.R. = Composite reliability; AVE = Average variance extracted.

4.3.2. International travel

For the international travel dataset, the same process was taken to analyze the data. First, CI was tested before the CR, convergent and discriminant validity of the model were examined. The same three items were excluded because factor loadings were lower than 0.7: an important country (0.605), a politically influential country (0.618), and a country that respects the environment (0.672). This second-order construct had both good convergent and discriminant validity. The standardized coefficients of the remaining 12 items ranged from 0.701 to 0.967. Table 3 shows that CI had good CRs and AVEs. Additionally, the square root of a construct's AVE was larger than its correlation with other constructs: $\sqrt{{\text{AVE}}_{\text{F}1}}=0.880$, $\sqrt{{\text{AVE}}_{\text{F}2}}=0.896$, $\sqrt{{\text{AVE}}_{\text{F}3}}=0.768$. This second-order construct also had a satisfactory model fit: ${\chi }^2$ = 420.231 (df = 43), ${\chi }^2$/df = 9.773, p < 0.001; GFI = 0.932, IFI = 0.970, NFI = 0.966, and CFI = 0.970; RMSEA = 0.094.

Table 3.

Correlation matrix of country image factors.

Constructs	F1	F2	F3	Mean	SD	CR	AVE
F1	1			4.808	0.196	0.945	0.775
F2	0.858	1		5.095	0.124	0.942	0.803
F3	0.609	0.665	1	5.555	0.170	0.804	0.590

Note: S.D. = Standard deviation; C.R. = Composite reliability; AVE = Average variance extracted.

Second, the model including CI, SK, DT, TA, and VI1 was investigated. The CRs and AVEs of the five latent constructs were larger than the cut-off points, indicating good CR and convergent validity (see Table 4 ). The square root of a construct's AVE was larger than its correlation with other constructs: $\sqrt{{\text{AVE}}_{\text{CI}}}=0.875$, $\sqrt{{\text{AVE}}_{\text{DT}}}=0.963$, $\sqrt{{\text{AVE}}_{\text{SK}}}=0.861$, $\sqrt{{\text{AVE}}_{\text{TA}}}=0.964,\text{ and }\sqrt{{\text{AVE}}_{\text{VI}1}}=0.930$, suggesting the discriminant validity was achieved. The indices revealed that this model had a good fit: ${\chi }^2$ = 1974.755 (df = 354), ${\chi }^2$/df = 5.578, p < 0.001; GFI = 0.901, IFI = 0.961, NFI = 0.953, and CFI = 0.961; RMSEA = 0.068.

Table 4.

Correlation matrix of the five constructs.

Constructs	CI	SK	DT	TA	VI1	Mean	SD	CR	AVE
CI	1					5.091	1.301	0.907	0.766
SK	−0.046	1				4.748	1.528	0.895	0.741
DT	0.791	−0.175	1			3.973	2.087	0.987	0.928
TA	0.719	−0.214	0.827	1		4.027	1.984	0.985	0.930
VI1	0.154	−0.158	0.152	0.232	1	3.797	1.972	0.950	0.864

Note: CI = country image; DT = destination trust; SK = subjective knowledge; TA = tourist attitude; VI1 = visit intention within one year; S.D. = Standard deviation; C.R. = Composite reliability; AVE = Average variance extracted.

4.4. SEM results

4.4.1. Domestic travel (within 1 year)

The relationships between CI, SK, DT, TA, and VI1 were examined for domestic travel of American and Canadian respondents. The significance levels of the coefficients of three paths (H4, H5, and H8) were greater than 0.05: H_{3 CI-VI1} = 0.435, H_{6 SK-VI1} = 0.720, and H_{8 DT-VI1} = 0.085, leading to their exclusion. All the remaining path coefficients were less than 0.001 (see Fig. 1 ), suggesting that these hypotheses (H₁, H₂, H₄, H₅, H₇) were supported. The model also had a good fit: ${\chi }^2$ = 1583.185 (df = 358), ${\chi }^2$/df = 4.422, p < 0.001; GFI = 0.899, IFI = 0.964, NFI = 0.954, and CFI = 0.964; RMSEA = 0.059. The results revealed that CI, SK, and DT would not directly impact domestic travel intention within one year under the adverse conditions of an ongoing pandemic, while all three significantly influenced TA, which further impacted on VI1. The direct impact of CI on TA was 0.216, while the indirect impact of CI on TA through DT was 0.296. This indicates that DT is a critical mediator to the relationship between CI and TA. DT is also a full mediator for the relationship between SK and TA. When people had more knowledge of the pandemic, they were more cautious and trusted the destination less, which influenced their TA.

Fig. 1.

Domestic travel (1 year)

***p < 0.001; n = 1000.

4.4.2. International travel (1 year)

The relationships between CI, SK, DT, TA, and VI1 of the international travel dataset were examined for American and Canadian respondents. The coefficients of two paths were not significant and excluded: (H_{3 CI-VI1} = 0.435 and H_{8 DT-VI1} = 0.053). The remaining paths had significant coefficients (see Fig. 2 ), indicating that these hypotheses (H₁, H₂, H₄, H₅, H₆, H₇) were supported. The model fit was satisfactory: ${\chi }^2$ = 1982.301 (df = 357), ${\chi }^2$/df = 5.553, p < 0.001; GFI = 0.883, IFI = 0.961, NFI = 0.952, and CFI = 0.961; RMSEA = 0.068. The model showed that CI and DT do not directly influence international travel intention within 1 year, while significantly impacting TA, which affects VI. The direct impact of CI on TA was 0.199, much smaller than its indirect impact through DT (0.516). Thus, DT plays an even more important role in influencing TA for international (see Fig. 2) compared to domestic travel. The indirect impact of SK on VI through TA was only −0.019, while the direct impact was −0.114, suggesting that when people had more knowledge of the pandemic, they were less likely to take an international trip within a year. This path was not significant for the domestic travel model.

Fig. 2.

International travel (1 year)

***p < 0.001; n = 1000.

4.5. Multigroup analysis

In addition to testing the nine proposed hypotheses, a multigroup analysis was also conducted to examine the differences between Americans and Canadians. To have a holistic view of people's travel intention within different time frames, the visit intention within 1 year (VI1) and within 2 years (VI2) were investigated separately. As the models have been confirmed with the convergent and discriminant validity (Table 2, Table 4), the mean values were used to represent each latent construct. For the model fit, GFI, IFI, NFI, and CFI were all greater than the cut-off points 0.9, and RESEA was less than 0.08.

4.5.1. Domestic travel (1 year)

The differences between Canadians and Americans for domestic travel within one year were tested by comparing the unconstrained and constrained models in AMOS. In the nested model comparison, when all structural weights were constrained, chi-square was 58.997 and p-value was 0.000, indicating that at least one significant difference existed. Therefore, two models were freely estimated except constraining one path to be equal across groups. Chi-square analysis showed that one relationship was moderated by nationality (Table 5 , Left). The effect of SK on DT for Canadians was not significant (p = 0.237 > 0.05), while it had a significantly negative impact on Americans' perceived DT of the US (p < 0.001). It indicated that although Canadians knew the severity and danger of COVID-19, this did not significantly influence their trust in Canada, because they believed their country had implemented appropriate control measures. In contrast, Americans' knowledge of the pandemic had a significantly negative effect on their trust in the US. The much higher case numbers and death rate in the US suggested that this country did not effectively handle the pandemic at the time of the study. Under a relatively high-risk environment, SK significantly impacts people's DT even for their own country.

Table 5.

Multi-group analysis for 1-year timeframe.

Paths		Domestic					Cross Border
		Canadians in Canada		Americans in US		Nested model	Canadians to US		Americans to Canada		Nested model
		Std. coefficient	P	Std. coefficient	P	P	Std. coefficient	P	Std. coefficient	P	P
H1	CI → DT	0.614	***	0.573	***	0.106	0.497	***	0.612	***	0.804
H2	CI → TA	0.106	*	0.209	***	0.176	0.158	***	0.166	***	0.909
H4	SK → DT	−0.042	0.237	−0.341	***	***	−0.321	***	−0.074	*	***
H5	SK → TA	−0.103	*	−0.167	***	0.248	−0.065	*	−0.167	***	0.268
H6	SK → VI1	n.s.	n.s.	n.s.	n.s.	n.s.	0.023	0.621	−0.214	***	***
H7	DT → TA	0.343	***	0.494	***	0.644	0.654	***	0.454	***	0.084
H9	TA → VI1	0.337	***	0.328	***	0.484	0.010	0.833	0.404	***	***

Notes: *p < 0.05, **p < 0.01, ***p < 0.001; CI = country image, SK = subjective knowledge, DT = destination trust, TA = travel attitude.

VI1 = visit intention within 1 year. n.s. = non-significant.

4.5.2. International travel (1 year)

Canadians and Americans had significant differences (chi-square = 129.604, p < 0.001) for international travel within one year, suggesting that nationality was an influencing factor for this model (Table 5, Right). The negative effect of SK on the perceived DT of the US by Canadians was significantly stronger (p < 0.001) than its effect on the perceived DT of Canada by Americans (r_Canadians = −0.321, r_Americans = −0.074). Canadians’ travel intention to the US was not significantly impacted by SK nor TA (p > 0.05) as they perceived high risks in travelling to the US in spite of a positive travel attitude. However, Americans' travel intention to Canada was significantly influenced by SK (r = −0.214, p < 0.001) and TA (r = 0.404, p < 0.001). The findings suggest that when people travel from what may be perceived as a relatively low-risk destination (e.g., Canada) to what is perceived as a high-risk one (e.g., the US), their knowledge of risk leads to significantly less trust in the latter, reducing visit intention significantly. In the reverse situation, people still intent to visit the destination, but the greater their SK, the more their visit intention decreases.

4.5.3. Domestic travel (2 years)

The relationships between the five constructs for the long-term (2 years) was also investigated. The differences between Canadians and Americans for domestic travel were tested by comparing the unconstrained and constrained models in AMOS (Table 6 , Left). The model was significantly different for Canadians and Americans (chi-square = 59.231, p < 0.001). Similar to VI1, the impact of SK on DT for Canadians was not significant (p = 0.237 > 0.05), while it had a significantly negative impact on Americans’ perceived DT of the US (p < 0.001). Interestingly, different from the model of VI1, CI was a significantly positive factor influencing VI2 (r_Canadians = 0.113, p < 0.05; r_Americans = 0.207, p < 0.001). It indicated that CI plays a less important role for travel intention within a short-time of the outbreak of a negative event, like COVID-19. However, it is still a critical factor for the development of a destination in the longer term.

Table 6.

Multi-group analysis for 2-year timeframe.

Paths		Domestic					Cross Border
		Canadians in Canada		Americans in US		Nested model	Canadians to US		Americans to Canada		Nested model
		Std. coefficient	P	Std. coefficient	P	P	Std. coefficient	P	Std. coefficient	P	P
H1	CI → DT	0.614	***	0.573	***	0.106	0.497	***	0.612	***	0.804
H2	CI → TA	0.106	*	0.209	***	0.176	0.158	***	0.166	***	0.909
H3	CI → VI2	0.113	*	0.207	***	0.775	n.s.	n.s.	n.s.	n.s.	n.s.
H4	SK → DT	−0.042	0.237	−0.341	***	***	−0.321	***	−0.074	*	***
H5	SK → TA	−0.103	*	−0.167	***	0.248	−0.065	*	−0.167	***	0.268
H6	SK → VI2	n.s.	n.s.	n.s.	n.s.	n.s.	−0.052	0.263	−0.161	***	**
H7	DT → TA	0.343	***	0.494	***	0.644	0.654	***	0.454	***	0.084
H9	TA → VI2	0.263	***	0.230	***	0.295	0.028	0.549	0.420	***	***

Notes: *p < 0.05, **p < 0.01, ***p < 0.001; CI = country image, SK = subjective knowledge, TA = travel attitude, DT = destination trust.

VI2 = visit intention within 2 years.

4.5.4. International travel (2 years)

Travel within 2 years showed three significant differences for Canadians and Americans (chi-square = 124.444, p < 0.001) (Table 6, Right): the impact of SK on DT (p < 0.001), SK on VI2 (p < 0.01), and TA and VI2 (p < 0.001). SK had a stronger effect on DT for Canadians travelling to the US than Americans travelling to Canada. Additionally, VI2 for Americans to Canada was significantly influenced by SK and TA. It appears that the different timeframes did not impact the relationships among the constructs.

5. Discussion and conclusions

This study set out to investigate the effects of CI, SK, and DT and their relationship to travel attitude and visit intention during COVID-19. This research used Canada and the US as cases to investigate people's domestic and international travel intentions. These two countries had significantly different approaches to handling the breakout of the COVID-19 pandemic. It tested the TA and VI of domestic residents and international travellers, specifically Canadians travelling domestically versus Americans travelling domestically, and Canadians travelling to the US versus Americans travelling to Canada. Two timeframes (i.e., one year vs. two years) were examined to further reveal the underlying relationships between those important constructs.

First, this study supports findings that CI positively influences travel attitude (Chung & Chen, 2018), and that DT is more prominent compared to CI under the adverse conditions of an ongoing pandemic. The indirect impact of CI through DT on travel attitude is significantly greater than the direct impact of CI. This research also contributes to understanding the roles of SK and DT in impacting travel attitude and visit intention, which have not been adequately investigated. Second, it confirms that SK of COVID-19 negatively influences both domestic and international travel attitude. However, the impact of SK on DT varies with the risk of the domestic environment. For example, although Canadians are aware of the severity of the pandemic, their knowledge does not influence their trust in Canada as a domestic travel destination whereas when Americans have more knowledge of COVID-19, they trust the US less. This suggests that Canadians, contrary to Americans, were not concerned with the handling of the pandemic in their country. This could be explained in part by Canada's lower number of cases per million people (2,753) compared to the U.S. (7774) and the fact that infections were highly concentrated in long-term care homes which registered 82% of deaths during the first two waves (Bejan & Nikolova, 2020). The significant negative impacts of SK on DT on international travel exist for both Canada and the US. Third, there were no direct impacts of CI, SK, and DT on domestic VI1. This finding differs from previous research (Chaulagain et al., 2019; Han & Hyun, 2013; Phillips et al., 2013), suggesting that travel attitude plays a critical role in leading to visit intention. During the pandemic, the perception of a destination and people's knowledge will influence their travel attitude but not necessarily their visit intention. Travel attitude is a full mediator implying that changing travel attitude is critical for the recovery of domestic travel. Interestingly, CI had a direct impact on domestic VI2, indicating that CI is still a crucial construct influencing travel intention, but its importance in the short-term decreases under an adverse event.

Additionally, it was found that despite a positive attitude toward the US, their travel intention was extremely low, likely because the pandemic situation in the US was so much more severe than in Canada at the time of data collection. Canadians seem to concur with their American neighbours about the poor pandemic response in the US, while Americans place a significant amount of trust in Canada as a destination whether in the one-year or two-year scenario.

The implications for destination management organizations are clear: trust in the destination and its handling of adverse events is paramount as is good communication around the event and measures taken to address it. Although such events might make potential tourists less trusting in the destination, it does not impact their domestic TI. Over the longer term, CI regains its influence over the decision to travel domestically although not internationally.

With international borders still largely restricted, if not closed, countries are increasingly focusing on domestic travel to aid the recovery of their tourism sector. Critical in that endeavour is a cohesive marketing message, especially around any COVID-19 restrictions and safety measures. Unfortunately, the government structure in both subject countries makes this very difficult as the federal level can only provide guidance and its jurisdiction is limited to national and international transportation, border crossings and entrance requirements, and mandates for its own personnel. This means that individual provinces and states adopt their own rules which can diverge greatly from the advice at the federal level. The tourism industry's repeated requests for consistency in measures enacted across the country and leadership in terms of developing some type of proof of vaccination have so far gone unanswered, although the Canadian government is said to be working with the international community on a vaccine passport (Major, 2021).

At the time the survey was conducted, both neighbouring countries were seeing a “flattening of the curve” of the pandemic's second wave. In Canada, almost all provinces/territories had either formed a travel bubble (in Atlantic Canada) or removed restrictions discouraging travel between provinces. In the USA, the few states that had imposed interstate travel restrictions had also largely rescinded these. In both countries, international travel for non-essential reasons was strongly discouraged, if not prohibited. This positive development is likely to have influenced attitudes towards travel somewhat, especially in the USA where over half the population declared themselves ready to travel again (Destination Analysts, 2020), albeit by car rather than by air. In spite of this, resident confidence in opening up communities to visitors remained subdued, although more positive in the USA than Canada (Destination Analysts, 2020; Twenty31, 2020). But marketing of destinations had to be done with great care as only about a third of respondents to these sentiment studies declared themselves ready to see advertising. Especially in Canada, advertising by international destinations is still being frowned upon even a year later (Destination Canada, 2021). The conclusions to be drawn from this and other research confirm that to build confidence, destinations and operators will need to build trust and strengthen community relationships. Residents need to be engaged and informed of the actions being taken to protect the health of their communities.

6. Limitations and future research

As mentioned, these data were collected in early August 2020. While this provides a snapshot of the importance of the constructs of CI, SK, and DT and their relationship to travel attitude and visit intention, it would be important to repeat the study now that a significant portion of the populations in both countries are vaccinated and they have emerged from a third and fourth wave of the virus driven by the spread of the highly contagious Delta and Omicron variants. Vaccine mandates are now imposed by both countries for federal employees with strong encouragement to large employers to do the same. Thus, it would be important to repeat the study at another time to gain a full picture of the impacts of the constructs on travel attitude and visit intention as the pandemic progresses. Along with DT, perceived risks might provide an additional explanation for travelers’ attitude and intention and should be considered in future studies. It would also be important to replicate the research in other jurisdictions such as Europe and Asia where the response to the pandemic and measures taken to combat the SARS-Cov-2 virus differed greatly. Although the loss of life caused by the pandemic continues to be heart-wrenching and tourism as an economic sector has suffered greatly, it is also a unique opportunity to study a long-lasting adverse event and gain deeper insight into the relative importance and interaction between the five constructs under investigation.

Appendix A. Supplementary data

The following are the supplementary data to this article: