*4.1. Exploratory Factor Analysis (EFA) and Reliability Tests*

CFA methodology recommends verifying the viability of the proposed model through EFA [36]. We, therefore, confirmed the relationships among the variables—first, to determine whether the proposed model is identified and, second, to verify the factor loadings on each of the variables. Table 4 presents the results.

**Table 4.** Consistency and internal validity, average variance extracted (AVE), and composite reliability.



**Table 4.** *Cont.*

Note: Factor loadings should be above 0.7, AVE values above 0.5, and CR scores above 0.7.

The factor loadings show consistency among the factors observed and the variables, except for item 3 (sales volume), which could be collinear with other variables and whose factor loading is below the accepted minimum of 0.7 [88]. We, therefore, excluded item 3 from the model. The other items and factors show satisfactory factor loadings, demonstrating the model's internal consistency. All values over 0.5 were accepted for average variance extracted (AVE) [36] and these values range from 0.64 to 1.00. Composite reliability (CR) values are between 0.75 and 0.99—in all cases, above 0.7, indicating construct validity [82]. Next, we present the adjusted empirical model (Figure 2).

The following values were obtained in the adjusted model: Chi-square = 363, degrees of freedom =126, standardized Chi-square = 2.88, root mean square error of approximation (RMSEA) = 0.021, Tucker–Lewis index TLI = 0.951, and incremental fit index IFI = 0.960 (these two indices should be >= 0.90). The indices for goodness of fit are incremental fit CFI = 0.960 and parsimony fit NFI = 0.940. According to theory, these values confirm the model's internal consistency, causal relationships among the variables, and good fit [82].

Next, the fact that the average of our six factors is associated with ICTS indicates that tourism business owners observed a high impact from the COVID-19 crisis. The tourism SMEs adopted remote work to reduce the impact on employment in the months of lockdown but could not sustain this measure economically over time. The measurement variable public aid to maintain payroll was crucial during the most critical months.

Item 3 (sales) in 2020 compared to 2019 measured the economic impact most precisely, showing a fall in income of 50% in over half the businesses that remained active; this calculation was obtained from the original revenue figures for 2019 and 2020, extracted from Orbis.

This item affected travel agencies in lower percentages. Many business owners had to assume the cost of reinvesting to reactivate their business to adapt to changes in technology and meet public-health requirements. Finally, the findings show that trends in both sales and payroll recovery in 2021 were factors determining whether firms were still affected by the pandemic compared to the most recent year under normal circumstances (2019). This factor strengthens indications that recovery is ongoing. Table 5 presents our evaluation of each hypothesis using the statistics obtained.

**Table 5.** Hypothesis contrast with structural model results.


#### **Table 5.** *Cont.*


Note: The regression weight obtained for the causal relationships corresponds to consolidation of the factors composing the observed endogenous variables relative to the unobserved variables.

**Figure 2.** Adjusted model.

*4.2. Business Situation (BS) When Facing the Economic Impact of the COVID-19 Crisis (ICTS)*

To test the validity of hypothesis 1 about how can the business situation of tourism SMEs in Colombia influences the level of economic impact caused by the COVID-19 crisis, we considered the weight of the regression obtained, 0.815 (Table 5), and the factor loading, 0.67. Although this factor loading is below the commonly accepted threshold of 0.7, we included it because it showed a statistically significant relationship to the criterion variable ICTS through three of the factors observed. Similarly, the t-value obtained in the hypothetical relationship is far from 0, validating the alternative hypothesis proposed, and a p-value below 0.05 permits us to reject the null hypothesis.

The economic subsector (item 1) obtained a factor loading of 0.71 (Figure 2) and a regression weight of 0.491—statistically significant values. Analysis of the percentage incomes for 2020 compared to 2019 shows that travel agencies suffered the greatest economic impact of the pandemic, with a loss of 58% income, followed by the lodging sector with 52%, and food and beverage with 40%. Tourism clubs lost 32%. We calculated these losses from the sales figures for 2019 and 2020, extracted from Orbis, classified by subsector. The percentages do not include firms that closed in 2020.

The link between number of workers (item 2) and firm size showed a factor loading of 0.77 and a regression weight of −0.676, demonstrating an inverse relationship. This finding indicates that, the larger the firm, the less severe the economic impact of the COVID-19 crisis. Sales volume (base year 2020) (item 3), did not, however, show a statistically significant factor loading. The reason may be the decrease in 2020 sales, which could exclude many companies from the SME category, as well as the incorporation into the market of new firms, for which we could not establish the degree of impact because they had no figures from previous years.

Main customers (item 4) produced a factor loading of 0.72 and a regression weight of 1.0. This result may be considered decisive for ICTs, as the statistical distribution of the variable highlights the fact that 65% of tourism SMEs sell their services primarily to consumers and families. Type of customer, thus, explains why the impact on sales was higher for these firms than for firms whose market focuses on other companies. Finally, the 11% that sell tourism services to public administrations experienced less impact from the pandemic.
