Study Area

The industrial sector is hugely important for the economic development of the state of Querétaro. In this sense, the study area is the PIBJ, located 11 km south of the geographic center of the Queretaro municipality and 4 km north of the center of Santiago de Querétaro. The PIBJ comprises an area of 450 hectares and started operations in 1963. Today, with nearly 60 years of operation, it continues supporting international and national companies that generate multiple job opportunities.

It is worth mentioning that the study area has been engulfed by urban growth in the municipalities of Querétaro, El Marqués, and Corregidora, which comprise the Querétaro Metropolitan Zone (QMZ). In addition, schools, households, and shopping malls have resulted in the Park being located in one of the areas with the highest public transport mobility in the state of Queretaro. The PIBJ is served by 31 different public transport routes (10, 12, 21, 27, 29, 31, 36, 40, 45, 50, 51, 61, 66, 67, 72, 76, 81, 84, 96, 98, 105, 110, 121, 130, 131, 132, 133, 134, 136, L07, and A1). The study area is shown in Figure 1.

Source: Own elaboration. SYMBOLOGY / Querétaro, Mexico / Queretaro Metropolitan Zone / Benito Juárez Industrial Park

Figure 1 Study area.  

Data Collection

Questionnaire Design

The questionnaire was structured into two main sections. The first is the Departure-Arrival Questionnaire (D-A) with multiple choice questions; the second is the questionnaire on the integral perception about COVID-19 with five-level Likert-type questions (not at all, low, moderate, high, very high), arranged in ascending order. The D-A questionnaire provides information related to the socio-demographic and commuting profile of users. The answers to the D-A questionnaire reveal the actual behavior of the respondent, that is, they regard revealed preferences. On the other hand, the questionnaire on the integral perception of COVID-19 provides information on the impressions generated by this disease, the perceptions of risk in different transport modes, and the understanding of sanitary measures in public transport. The answers to this questionnaire reveal the likely behavior of the respondent under hypothetical scenarios, i.e., they regard stated preferences.

The purpose of addressing socio-demographic characteristics is to identify how they ultimately affect whether users choose public transport in the face of the COVID-19 pandemic. Commuting characteristics were addressed to establish whether they are relevant when users choose public transport in the face of the COVID-19 pandemic. On the other hand, questions about the impressions generated by COVID-19 aimed to distinguish the knowledge, understanding, and perspective of users through the socio-psychological impressions produced by the pandemic. Asking about the likelihood of contracting the virus is useful to determine the perception of users about becoming infected under various hypothetical scenarios. Finally, the purpose of asking about the understanding of sanitary measures in public transport is to identify the users' perspective on the effectiveness of preventive measures, the established guidelines, and the compliance level with them.

Sampling Scenarios

The size of the population currently working in the PIBJ was determined by consulting the National Statistical Directory of Economic Units (Directorio Estadístico Nacional de Unidades Económicas; DENUE, in Spanish) (^{INEGI, 2021}). In this directory, data on the identification, location, economic activity, and size of active businesses were reviewed, mainly in the large-business segment. In this way, based on the data mentioned above, the number of workers in the PIBJ was estimated to be about 7500, distributed in 130 companies focused on the industrial growth of the state of Queretaro.

The sample size was established using Equation 1 (^{Ibeas et al., 2007}) with a 95% confidence level, a standard normal variable of 1.96 (corresponding to a 95% confidence level), and an accuracy level of 0.05 (maximum error of 5%):

Where n is the sample size surveyed; p is the proportion of trips to estimate (0.5 being the most conservative); Z is the standardized normal variable for the required confidence level; E is the accuracy level (expressed as a ratio); N is the total population size.

From Equation 1, the number of questionnaires necessary to be administered was 365.

Application Scenarios

Before administering the final questionnaires, a pilot questionnaire was applied to determine whether the survey would operate efficiently and whether the level of understanding of the questions was appropriate for the respondent. This approach allowed for reviewing the information obtained and making adjustments to the questionnaire, if necessary.

With the authorization of the local government in the study area, a total of 365 valid individual questionnaires were administered to users of PIBJ public transport from October to November 2021. To note, during the questionnaire application period, the situation of the pandemic in Mexico remained with a decreasing trend in confirmed cases. The third wave of coronavirus peaked in mid-August 2021; since then, there has been a sustained reduction in the number of infections in the country. In addition, the vaccination strategy positively influenced the pandemic indicators in Mexico; by the end of October 2021, 85% of the population over 18 years old had already received at least one dose of the anti-COVID-19 vaccine.

Following the established strategy and the schedule, approximately 20 to 30 questionnaires were applied per working day between 09:00 and 12:00 hours. The steps to administer the questionnaires were:

It should be mentioned that the questionnaire was administered only to users of PIBJ public transport, as required by the study. Potential respondents were selected by asking if they were public transport users during the COVID-19 pandemic or previously. This avoided collecting invalid questionnaire information.

Theoretical Background

Data Analysis Method

The binary logistic regression model was used to analyze the influence of the explanatory variable on the dependent variable when the latter involves two options. This model uses the maximum likelihood estimate when the dependent variables are transformed into logit variables (^{Ortuzar and Willumsen, 2008}; ^{Koppelman and Bhat, 2006}).

Thus, the binary logistic regression analysis is part of generalized linear models and is characterized by its response distribution (binomial) and a binding function (^{Dalgaard, 2008}). The binary logistic model is expressed as shown in Equations 2 and 3:

Where P is the probability that Y = 1; x _i is the explanatory variable i; α is a constant; and β _i is the regression coefficient associated with the variable i.

For the present study, a binary logistic regression model was implemented to predict the result of a categorical variable (qualitative binary variable) dependent on the predictive or independent variables (explanatory variables). In this case, the dependent variable was the choice of public transport, where a Y = 1 means choosing public transport and Y = 0 means choosing not to use public transport.

Preliminary Analysis of the Variables Affecting Modal Behavior

Socio-Demographic Characteristics

Each public transport user has different socio-demographic characteristics, leading to a heterogeneous choice behavior between them. For this reason, when users made decisions about using public transport during the COVID-19 pandemic, they were affected by multiple socio-demographic characteristics. The descriptive statistics corresponding to the socio-demographic characteristics obtained from the questionnaires are shown in Table 1.

Of the 365 respondents in the sample, 73 (20.0%) were female and 288 (78.9%) were male. Since most public transport users working in the PIBJ are middle-aged persons representing the main labor force in society, most respondents were between 31 and 40 years old, representing 44.11% of the sample. Regarding the household role, our data showed that 158 respondents (43.29%) were the head of the household and 140 (38.36%) were sons/daughters of the household head. Regarding the completed education level, approximately 38.90% of users completed college education or higher, followed by 35.89% who completed high school. As for occupation, 46.30% of respondents had a low-level job (cleaning, driver, assistant, and security), and 35.07% had an intermediate-level job (technical/administrative); given the nature of their jobs, these respondents needed to use public transport during the COVID-19 pandemic. Regarding income, 71.51% of respondents earned $10 000-$15 000 Mexican pesos per month, and 20.27% earned less than $10 000 per month. These results were to be expected, as higher wages are related to the possibility of traveling in a more expensive mode than public transport.

Commuting characteristics

The main commuting characteristics are the distance from the household to the workplace, walking time from the household to the public transport, commuting time, transport cost, weekly commuting frequency, and transport mode before and during the COVID-19 pandemic. These characteristics were undoubtedly affected when users chose to use/not to use public transport. Table 2 shows the descriptive statistics obtained from the questionnaires regarding the commuting profile of users.

Regarding the distance from the household to the workplace, 47.67% of respondents travel between 5 and 10 km, i.e., they live relatively close to the workplace, considering that the average linear size of the city of Querétaro is 15 km. Concerning the walking time from the household to public transport, 54.79% of respondents walk between 5 and 10 minutes - an acceptable walking time, considering that a person can walk 1 km in 15 minutes, on average. Besides, average commuting time is longer than 30 minutes for 46.03% of workers; considering the heavy traffic in the city of Queretaro, moving across the city commonly takes more than 30 minutes. The travel cost reported by 58.63% of the total respondents is between $10 and $20 Mexican pesos per trip. This cost is relatively acceptable since the average travel cost in urban areas in Mexico is between $6 and $14 Mexican pesos, which encourages users to travel by public transport. With regard to the frequency of commuting per week, 53.15% traveled 3 to 4 times a week to their workplace. This means that remote work from home increased during the pandemic due to restrictions mandated by the authorities; normally, the private sector works 5 or 6 days per week. As for the travel mode used by respondents during the COVID pandemic, 82.19% of users reported that they continued using public transport as their travel mode. These numbers indicate that 17.81% of users stopped using public transport and opted for another travel mode.

Integral Perception of COVID-19

The integral perception of COVID-19 in this study considered three main aspects included in the questionnaires administered to users: the impression of COVID-19 in public transport users, the perceptions of risk in different transport modes, and the understanding of the sanitary measures implemented in public transport.

The integral perception level about COVID-19 was graded according to five Likert levels (not at all, low, moderate, high, very high), where level 5 reflects the highest perception level regarding the impression about COVID-19, the risk of contracting the virus, and the sanitary measures implemented by the authorities in public transport. On the other hand, level 1 is the lowest perception level regarding the impressions about COVID-19, the risk of contracting the virus, and the sanitary measures implemented by public transport authorities.

The impressions about COVID-19 were mainly given grade 4 (high), specifically the level of understanding of the pandemic (44.66% of respondents selected grade 4), level of concern about the pandemic (37.26%), level of discomfort about the pandemic (43.56%), level of threat to life from the pandemic (48.77%), and level of awareness about the consequences of becoming infected (53.34%). These figures indicate that most respondents have strong negative impressions about the COVID-19 pandemic but, at the same time, have a high level of understanding and knowledge of the situation. The descriptive statistics corresponding to the impressions about COVID-19 are shown in Figure 2.

Source: Own elaboration.

Figure 2 Descriptive statistics of the impressions about COVID-19. 

As for the perception of risk in different transport modes, the possibility of becoming infected from public transport, and whether the responder considered it risky to use public transport, most scores corresponded to grade 4 (high), representing 66.30% and 52.33% of all answers, respectively. Besides, the possibility of becoming infected from public transport and from traveling by car was scored with grade 3 (moderate), representing 67.67% and 51.78%, respectively. The possibility of becoming infected by cycling and by walking was scored mainly with grade 2 (low), representing 45.21% and 41.92%, respectively. The above indicates that public transport users perceived the risk of using public transport in the face of the COVID-19 pandemic and considered that using an alternative urban mobility mode was less risky. The descriptive statistics corresponding to the possibility of catching the virus are shown in Figure 3.

Source: Own elaboration.

Figure 3 Descriptive statistics on the perceptions of risk from different transport modes. 

The understanding of the sanitary measures applied to public transport were graded 4 (high), specifically regarding the level of acceptance of sanitary measures (58.63% of all answers were given grade 4), level of compliance with sanitary measures (48.22%), whether the use of a face mask is considered appropriate (44.11%), whether social distancing is considered appropriate (43.01%), whether avoiding crowds is important (47.40%), and whether cleaning the transport units is important (47.67%). These numbers indicate that public transport users accepted, complied with, and understood the sanitary measures implemented by public transport authorities in the face of the COVID-19 pandemic, aiming to protect their health and life. Clearly, users perceived public transport as a high-risk space for the spread of the virus. The descriptive statistics corresponding to the possibility of catching the virus are shown in Figure 4.

Source: Own elaboration.

Figure 4 Descriptive statistics on the understanding of sanitary measures. 

Analysis of the Binary Logistic Regression Model

With the results of the questionnaire variables, a logistic regression model was constructed to predict the result of the dependent (categorical) variable based on its interaction with the explanatory (numerical-categorical) variables. Thus, the explanatory variables were the socio-demographic characteristics, travel characteristics, and the integral perception of COVID-19.

To select the best model, i.e., the one yielding the best results, a series of possible binary logistic models were constructed with the most significant variables included in the questionnaire. Based on the significance of the variables in the models analyzed, the following variables were deemed significantly related to the modal shift of users when choosing public transport during the COVID-19 pandemic: occupation, commuting time, level of knowledge of the consequences of being infected, possibility of being infected by public transport, possibility of becoming infected by means other than using public transport, and whether the user considered it risky to use public transport. The significant variables mentioned above are listed in Table 3.

The estimated logistic model with better characteristics in terms of the significance of its variables and prediction level is shown in Table 4. This model considers two groups of users: the first includes users that chose to use public transport; the second, users that chose not to use public transport. The estimated linear equation from the results in Table 4 is shown in Equation 4:

Table 4 shows that the p-value, corresponding to Wald’s significance test of the coefficient of each explanatory variable, is less than the maximum significance level of 0.05, indicating a significant linear relationship between the explanatory variable and Logit P (^{Koppelman and Bhat, 2006}; ^{Hilbe, 2009}).

In addition to Wald’s significance test, we also performed an Analysis of Variance (ANOVA). This analysis is used to compare the decrease in residual deviation when new variables are introduced into the model (^{Dalgaard, 2008}). Table 5 shows (ANOVA) a significant model improvement after each variable is included. Since all variables are significant (p < 0.05), they cannot be removed from the model. The tests in the ANOVA table are not completely independent, but the approximation should be good.

Other important tests for the interpretation of results are the model goodness-of-fit (pseudo R ²) since they measure the increase in the likelihood of the model explained by the independent variables of the regression model. These tests are used for assessing model fitness. In this case, a value closer to 1 means that the model is more significant or has a perfect fit; however, values close to 0.4 can be considered an excellent or satisfactory fit (^{Ortuzar and Willumsen, 2008}). McFadden's goodness-of-fit test is analogous to R ² in the linear regression model (^{Hilbe, 2009}). Therefore, Table 6 shows that the model coefficient is acceptable (in terms of prediction), given that the McFadden value is 0.287. Similarly, the Hosmer-Lemeshow value is 0.287. In addition, the Cox-Snell and Nagelkerke values are 0.236 and 0.388, respectively, suggesting that the model fit is acceptable.

On the other hand, the overall model accuracy model expresses the consistency between the results predicted by the model and those observed in the selection behavior. Since the model outputs are probabilities and the actual results are binary variables, if the probability is greater than 0.5, a result equal to 1 is considered, for probability values below 0.5 are assigned a result equal to 0. The model attained a data fit accuracy of 85.47%; this value means that the model is good, and values close to 100% indicate greater model suitability. We can conclude that the results obtained from the binary logistic model are satisfactory. The summary of the tests and the overall accuracy of the model are shown in Table 6.

This model can provide approximate estimates of the distribution of passenger flows in various transport modes in the city and provides data suitable for supporting traffic control plans and emergency measures developed in advance by traffic management authorities.

The results show that occupation and the underlying factor, i.e., related during the pandemic, were statistically significant since respondents with low-level jobs (low monthly income) have a odds ratio to continue using public transport 0.399 higher relative to respondents with high-level jobs (high monthly income). This is because high-income people often own a private car and would prefer to use it regularly regardless of the circumstances.

Our results also show that with longer commuting times within units, users have odds ratio up to 26.032 higher to continue using public transport, given the high cost of traveling on another transport mode. In addition, users who are more aware of the consequences of becoming infected have odds ratio 0.500 lower to continue using public transport than users who are less aware of the situation. Similarly, users who consider that the possibility of becoming infected is high while using public transport and those who consider it risky to use public transport have odds ratios to continue using it (0.112 and 0.555, respectively). Finally, those who consider that the possibility of becoming infected is the same in either public transport or any other travel mode has odds ratio 3.208 higher to continue using the former.