what will it take for students to feel safe to return to campus? Free

The COVID-19 pandemic has impacted schools and universities worldwide (“Impact,” 2021). The pandemic has forced total closure or near total closure of educational facilities at all levels. The goal of these closures has been to reduce and slow the spread of COVID-19. As of January 2021, it was estimated that over 800 million students worldwide had been affected by these closures (“COVID-19,” 2020; Skulmowski & Rey, 2020). According to UNICEF (2020), 23 countries were implementing nationwide closures with 40 countries implementing local closures. Taken together,these measures affected approximately 47% of the world’s student population.

These school closures not only impacted students, teachers, and their families (“May 2020 Examinations,” 2020), but they have also had far-reaching economic and social consequences (“Adverse Consequences,” 2020; Bao et al. 2020; Lindzon, 2020). Pandemic-driven school closures have also surfaced a number of economic and social issues, including: digital learning (“Distance Learning Solutions,” 2020; Lindzon, 2020), student debt (Aris- tovnik et al., 2020), food insecurity (Karp & McGowan, 2020), and homelessness (“Schools Race,” 2020; Sessions, 2020), as well as access to housing (Feuer, 2020), health care (“Coronavirus Forces Families,”2020), childcare (Ngumbi, 2020), the internet (Barrett, 2020), and disability services (“Education Dept. Says,” 2020; Jordan, 2020). The impacts have been more severe on disadvantaged students and their families (COVID-19,“ 2020).

For these reasons and many others, it is imperative that schools reopen as soon as possible. At the university level, getting students back on campus is the top priority. But, how will campuses keep students safe? Do students feel safe returning to campus? What precautions do universities need to take to help students feel safe when they reopen?

This article reports on a survey taken about student attitudes toward returning to campus and the safety precautions universities can and should take to keep students safe. The high response rate in the survey (3,940 responses from a total of 21,593 students) supported in depth demographic analyses of the responses by race and ethnicity, gender, age, education level, and academic disciple. This was coupled with a latent class model showing 5 distinct groups of students and their requirements for returning safely to campus. The article is organized in the following sections: a literature review of issues associated with the impact of school closings and other related work; the survey methodology; results on the demographic analysis; the latent class analysis model and discussion of the results; and conclusions and planned future work.

The research reported in this article focuses on identifying mitigation strategies universities could employ to improve student safety on campus as well as the conditions under which students would feel safe returning to campus during and post pandemic. This research provides insight into “how” reopening university campuses might be achieved and “what” actions university decision makers could take to attract more students back to campus. This work was done as part of a larger study seeking to understand student preferences toward online education.

The main contributions of this work lie in its new insights regarding the value students place on COVID-19 mitigation strategies and the affirming of results found by other researchers. The work highlights what positive actions campuses can and should take to enable students to feel safe and thereby want to return to campus rather than to simply continue online or switch universities. In addition, our research provides a detailed demographic analysis and utilizes a latent class model to identify five specific groups of students that can be applied in other settings to help understand student preferences and the dynamics behind student behavior pertaining to decisions about returning to campus.

To protect student health and safety during the COVID-19 pandemic, students at all levels in many countries were sent home and most education was continued online (“COVID- 19,” 2020). Much research is underway to assess the severe economic, social, and mental health costs of the pandemic to students, faculty, and staff (Biber et al., 2020; Jones et al., 2021) as well as to understand what can be done to address these impacts (Arenas et al., 2020).

For instance, research done by the Yale school of medicine (Halperin et al., 2021) shows that anxiety is up significantly among college students over prepandemic levels. Researchers at Tulane University have found that students who lack access to university services experienced “increasing housing and food insecurity, financial hardships, a lack of social connectedness and sense of belonging, uncertainty about the future, and access issues that impede their academic performance and well-being” (Lederer et al., 2020). Other researchers have documented the impact of the pandemic on student performance (Engzell, 2021). While others describe a variety of mental health issues (Grubic et al., 2020).

Understanding student preferences and the mitigations universities can take to enable students to feel safe is essential to getting students to actually return to campus. There are a lot of online options available to students which enable them to complete their degree off-campus if they do not feel safe returning.

Given the magnitude of the impacts and the costs to global society on all levels there is widespread recognition that getting students back to school is extremely important. Some researchers, including those from the National Institute of Health (Losina et al., 2021) and others (Zafari et al., 2020), have taken a strictly financial view to making decisions about returning to campus. While the financial perspective is important, we believe that keeping students safe and understanding student preferences are essential aspects in making the decision regarding “when” to reopen to campuses and for developing plans as to “how” the return will be implemented.

How can returning to campus be done safely? The work reported in this article focuses on understanding the circumstances and set of mitigations that can be taken to enhance student willingness to return to campus at the university level.

Work done at the University of South Carolina (Edwards et al., 2020) shows that factors such as attitude toward vaccinations, risk of exposure and risk perceptions contribute to student willingness to come back to campus. Our study affirms these findings by showing that mitigation efforts on campus are just one of the factors contributing to students’ perceptions of on-campus health and safety.

Edwards, Francia, and Van Willigan at East Carolina University (Qiao et al., 2020) found that some students did not have access to certain resources (computers, internet access, etc.) and experienced increased responsibilities due to the pandemic. The research found that these other factors are important to students’ willingness to return. Our study affirms these findings as well as investigating other factors such as student perceptions towards their learning environment (i.e., in-person vs. online) that may influence students’; willingness to return.

Latent class modeling is the right approach when the group under analysis has several distinct groupings. Some successes for this kind of analysis are evidenced by Nagelkerke et al. (2016). As will be seen in the fourth and fifth sections, there was more homogeneity within the classes identified by our model than exists within typical demographic groups of race, gender or academic discipline.

This article seeks to provide university leaders with information and insights about student preferences that they can use to make decisions about how and under what conditions to return to campus. This article affirms many results found by other researchers while adding new insights regarding how students value COVID-19 mitigation strategies. We identify some of the strategies campuses can implement to enable students to feel safe and thereby want to return to campus rather than to simply continue online or switch universities.

In November of 2020, an online survey was sent out to all students currently enrolled at one midsized public university. We received 3,940 responses from a total of 21,593 students enrolled during fall term 2020 (an 18.2% response rate). Of these submitted surveys, 725 contained primarily missing data, and were deleted from the analysis, with responses to 3215 surveys available for analysis. Of these 3,215 valid sets of survey responses, 3,053 had no missing across all the items relevant for this analysis. An analysis of the representativeness of the various subgroups along each demographic category of concern is presented in the Appendix. The demographic categories of consideration in this article include: age, race/ ethnicity, gender, grade point average (GPA), student level and academic discipline. Table 1 shows subgroups that were overrepresented and underrepresents in the sample by demographic category.

As can be seen in Table 1, females and highGPA students were overrepresented in the sample, while males and students with GPAs below 3.0 were underrepresented. In terms of race/ethnicity, white students were overrepresented by 3% while black, Hispanic, international and unidentified students were each underrepresented by 1%. In terms of age, the 30-40 year old subgroup was disproportional at a level of +4%. Only subgroup differences of more than + or less than - 3% were included in Table 1.

There were several subgroups that were excluded from the analysis of their demographic category due to small number of respondents. These subgroups included: pacific islanders, students over 60 years of age, nonadmitted and post bac students, and students in graduate interdisciplinary studies. Further details comparing subgroup representation to the study population demographics can be seen in the Appendix.

The primary question analyzed in this 2020 survey was, “What COVID-19 mitigation strategies and conditions need to be in place before you, as a student, would feel safe returning to campus?” The results showed that while a small percentage (14.8%) of studentswould be interested in coming back to campus right away (without any added precautions), the majority of students would like to see a variety of safety measures and preconditions put in place prior to returning to in-person classrooms.

In terms of actions universities can and should take, the survey showed that students would like to see the following safety measures:

• special cleaning protocols between classes (66.4%);

• social distancing in the classroom (69.4%);

• mandatory mask wearing (72.8%);and

• temperature screening at the doors of classrooms (46.6%).

As a precondition for returning to in-person classes, the majority of students (64.3%) would like to see local infection rates reduced to levels prescribed by the governor/Centers for Disease Control and Prevention¹ for safe opening. Furthermore, many students would like to have the vaccine in place prior to returning to campus. Yet, there is a portion of students who are not sure if they will feel safe returning to campus any time in the near future and would like to complete their education online (26.9%). The latent class model Section 5 discusses patterns of student sentiment identifying five distinct groups. Section 4 provides an analysis of multiple demographic groups including gender, race and ethnicity, academic discipline, age, education level and others.

Beyond the analysis in Section 3, we analyzed survey data along a variety of demographic categories including: age, race/ethnicity, gender, student level and enrollment status, and instructional unit.

Table 2 shows the possible options students could select in the survey, the descriptions of these options, and the average support for each. Overall, there is high support for mitigations of classroom cleaning, social distancing, and wearing masks. Also, students are concerned about local infection rates.

Figures 1 through 8 present results from our demographic analysis. To better highlight the variation across subgroups, the figures show the differences between the subgroup response and the average responses from Table 2.For example, in Figure 1, the number of female students with GPAs less than 2.49 who feel“safe now“ and would like to return to campus is 5 out of the 61 or 8.2% of the students in this in this subgroup. However, only 14.8% of all respondents feel “safe now“ on average. The difference between the average (14.8%) and the percentage of female respondents in the GPA less than 2.49 group (8.2%) is -6.60% which is displayed in Figure 1 with a lighter shaded box indicating this subgroup has less support than average. In contrast, male students with similar GPAs have 15.97% more support than average for feeling “safe now“ and are depicted in darker shades. In other words 30.77% of this 52 person subgroup feels safe now (i.e., 16 respondents). The number below the percentage in each cell of the figures indicates the total number of respondents comprising the subgroup. Continuing the example, there are 61 respondents in the subgroup of female students with GPAs less than or equal to 2.49. This number (61) is shown in the boxes in the first column of Figure 1.

Furthermore, for all diagrams below, the x- axis describes the subgroup/demographic binning, the y-axis outlines the safety measures via a heat map featuring scaling using shaded bins. The results are presented in the following sections.

As can be seen in Figure 1, there is a significant divergence between the genders and their respective views on precautions to COVID. Overall, females tend to place a higher value on COVID safety measures than average and females are below the average in viewing campus as being safe now. Males, on the other hand, tend to view safety precautions with below average enthusiasm and have more confidence that it is safe to return to campus now without any additional safety precautions. Interestingly, there is only a small trend in both females and males that higher GPA students and new students prefer more precautions than their lower GPA counterparts.

Racial equity is an important issue at many universities. Interestingly, partitioning by race showed a large variability in the support or lack of support for certain mitigations. In this section, we identify differences in the perceptions of feeling safe to return to campus based on this demographic dimension. As can be seen from Figure 2, students identifying as Native American, African American, orBlack, and declining to respond were among those who felt the most safe returning to campus without any precautions.

Students who identified as Asian, multiple ethnic/race, and Native American were among the groups interested in the mitigation strategies at above average rates. Furthermore, students identifying as Asian were among the groups feeling least inclined to return to campus right now. At the same time, fewer international students than the average will “never feel safe” returning to campus.

Regarding special room cleaning, masking and social distancing, students identifying as African American or Black were consistent or had small deviations from the average; however, this group diverged more significantly from the average on three mitigations: vaccines, rates down, and temp screening. African American students supported temperature screenings but did not feel safe on campus purely from lower infection rates in the community or having a vaccinated community. fewer Hispanic/Latino students felt comfortable returning to campus right now than average (3% less than average). These students were more interested in the mitigation strategies of special cleaning, mandatory mask wearing, and temperature screening. Students identifying as White across the board appear to fall slightly below the average and showed less concern for COVID.

Our hypothesis was that younger students would feel more boldly about returning to the classroom without precautions and that older students would feel less inclined to return due to the severity of COVID in elderly patients. We also expected that older students would support vaccinations. The results presented in Figure 3 show something different. Consistent with the hypothesis, students under 20.5 years old were the most enthusiastic about returning to campus immediately without any safety measures put in place. However, the second age bracket that was most inclined to return to campus were those students who are over 50. Moreover, students over 30 consistently diverged from the average by having less interest in mitigation strategies of wearing masks, special cleaning, etc. being put in place before feeling safe to return to campus. Students over 50 were the least interested in vaccines (7.9% below average) or low infection rates in the community (18.5% below average). Equally as surprising, students in the 17.5 to 25.5 age brackets were the most interested in mitigations like special cleaning of classrooms, wearing masks, and temperature screenings at the door.

Figure 4 shows student reactions to the safety of returning to campus based on student level and enrollment status. Significant differences appear between graduate students plus seniors and juniors, sophomores and freshmen. For undergraduates, freshman support more mitigations and conversely have a larger population than average which feel safe on campus. Graduate students support social distancing and discount special cleaning, temperature screenings, and masks. This lack of mitigation support also shows in this group's lower than average support for returning to campus.

In this section, we examine the differences in preference based on academic discipline which can be used as a proxy for career path and professional differences. As can be seen, students in engineering and computer science as well as students in school of business veer from the average due to their high support for viewing returning to campus as being safe during the COVID pandemic and discounting the need/value of having mitigations put in place prior to that return. Across almost all mitigations, other than vaccine, the college of engineering and school of business were below the average support seen at the university. Notably, engineering students’; preferences tended to be significantly below the average on selecting “never return to campus” providing further evidence that this student demographic would like to return to in-person education as soon as possible.

Alternatively, students who were undeclared, attended the school of public health, attended the college of liberal arts and sciences, or were enrolled in the college of education tended to have more support than the average regarding the value/need for mitigations before returning to campus. Interestingly, public health students fell below the average in requiring vaccines before returning to campus, and instead found the mitigations of cleaning, distancing, lower rates, and temperature screenings as adequate procedures for a safe return.

Latent class analysis constructs a model that classifies the respondents into mutually exclusive groups, where each group corresponds to a latent class. Respondents classified into the same class share a similar response profile to the corresponding survey items. Different combinations of endorsed items correspond to different classes of respondents.

As mentioned in Section 3, a total of 3,940 surveys were obtained. Of these submitted surveys, 3,053 had no missing responses across all the items. Only complete responses were used for this analysis.

The goal is to extract as many classes as needed to adequately account for the data. At the limit, a separate class could be constructed for each unique set of responses leading to perfect fit, but that level of detail would likely overfit the resulting classification model to the idiosyncrasies of sample error. Instead, the best is to compromise between model parsimony and model fit.

Choosing the number of latent classes for a latent class analysis model follows from empirical analysis. Four fit indices are analyzed for each number of classes, which vary from two to nine. The first two indices are the related information criterion indices, Akaike’;s information criterion (AIC) and Bayesian information criterion (BIC), comparable on the same scale. The other two indices are the likelihood ratio/deviance statistic (G2), and the Pearson chi-square goodness of fit statistic (?²), which can also be compared on the same scale.

Figure 6 shows the results of these fit indices from two to nine classes. Smaller values indicate better fit.

Although multiple technical issues differentiate the construction of these fit indices, including for each index different assumptions and asymptotic distribution that forms the basis of the index, all four indices lead to the same conclusion regarding the number of classes to retain in the LCM. The balance between parsimony and fit leads to the specification of either four or five latent classes. Increasing the number of classes from two to three to four leads to a substantial improvement in fit for each step. Moving from four to five classes provides a more modest improvement than the previous increases. Moving from five to six classes either shows little improvement, or for the BIC index, worse fit.

The BIC index penalizes model complexity more than the others. The BIC index helps avoid overfitting a model to the data by adding complexity, here the number of classes, solely to achieve fit to the data from which the model is estimated without regard to generalization beyond the specific and arbitrary encountered sample. For these data, BIC indicates worse fit as the number of classes increases beyond six, and smaller improvement than the other indices moving from four to five and then five to six classes.

To facilitate the decision between four and five classes, consider the minimal number of respondents per class. For the five-class model the smallest class membership is 6% of the sample. Even though the four class model has one fewer class, the lowest class membership is slightly less, only 5%.

For these reasons, the five-class LCM was selected. The AIC, G2, and X2 fit-indices report better fit with more classes because homogeneity within a class can generally be increased with smaller class membership. Unfortunately, overfitting a model to the one sample from which it was estimated mitigates the improved fit.

Class assignment follows from the maximum posterior probability from the data for each respondent of class membership calculated across the five classes. The validity of the five-class solution is enhanced by the large number of classifications into the corresponding class with large probability values, as shown in the following histogram.

Eighty percent of respondents were classified into their corresponding group with a posterior probability of 0.90 or above. Ninety percent of respondents were classified with a probability of 0.80 or above, and 93% of all respondents were classified with a probability of 0.70 or above. Given that the sum of the five posterior probabilities for each respondent across the five classes is 1.0, these high maximum probabilities indicate that class membership for most respondents is unequivocal for the five classes applied to this sample of respondents.

To examine the relationship of class membership to other variables, a refined selection of the data was derived to include only those students who were unequivocally assigned to one of the latent classes. The chosen cutoff maximum probability is 0.7.

The remaining respondents define relatively homogeneous response profiles for each group. The reduced number of respondents is 2,992, a reduction of 223 respondents.

We define the meaning of each of the groups according to the conditional probabilities of endorsing each item by group membership. If a respondent is assigned membership to a specific class, what is the probability of endorsement of a specific item? For example, a member of Class 1 has a 0.98 probability of wanting masks mandatory, whereas a member of Class 5 only as a corresponding probability of 0.07.

Based on the pattern of item endorsements, the five classes were named as follows, ordered according to the degree of endorsed caution.

• Class 1: Maximum caution (37%): Need all cautionary possibilities.

• Class 2: Much concern (19%): Similar to the Maximum Caution group, with slightly less emphasis except no temperature screening.

• Class 3: Internal factors (19%): Need events that the university can control, though with less emphasis on temperature screening.

• Class 4: External factors (6%): Need events outside of the control of the university, low infection rate and effective vaccine.

• Class 5: Minimal concern (19%): No concerns.

To feel safe to return to campus requires the following events for each group, with the groups named and defined as follows. The largest group of students, 37%, requires maximum caution to feel safe, almost twice as large as the group of students with minimal concerns, 19%. The much concern group, 19%, shares the same concerns as the maximum caution group, except no need for temperature screening. Together these two groups consist of 56% of all respondents. The Internal Factors group, 19%, places much less emphasis on the overall infection rate, but still requires the university internal controls, though again with a reduced level of endorsement.

The smallest group of students, External Factors at 6% of respondents, does not rely upon precautionary actions taken by the university. This group of students relies upon external factors, requiring a low infection rate and an effective vaccine. They want society as a whole to be safe, precluding anything the university by itself can accomplish.

A substantial minority of students, 19%, expresses little endorsement of any of the specific cautionary items, either internally controlled by the university or as the state of society in general.

In this section, we discuss how class membership varies along different demographic category lines.

When comparing female responses to males, males demonstrate a significantly risk-

ier profile than females (?² = 41.75, df = 4, p value = 0.000). For females, 41% are classified into the maximum caution class, with another 20% in the much concern class, and only 17% in the minimal concern group. For males, 11% less are in the maximum caution class, 30%, and 7% more are in the minimal concern class, a total of 24%.

We examined the class composition for each self-identified racial/ethnic group. For this analysis, remove the 120 respondents who declined to respond to the race item. Composition by race significantly varies across the five latent classes (?² = 78.62, df = 28, p value = 0.000).

The races with the highest proportion of membership in the maximum caution group is Asian with 303 Asians or 10% of the entire sample. On the contrary, only 35% of Whites are classified into the maximum caution class, with 20% classified into the minimal concern class. This compares to only 14% of Asians classified into the minimal concern class.

Blacks have a similar profile to Whites, with 43% classified into maximum caution, but 20% classified into the minimal concern class. The Hispanic/Latino classifications are between Asians and those of Whites and Blacks with 45% classified into maximum caution and 18% classified as minimal concern.

One distinction is the distribution of Asians across the five classes. Although comprising 9% of all respondents, they have a percentage of membership in the maximum caution group of 56%. The corresponding pattern for His- panic/Latinos, who comprise 14% of the total sample, is not as differentiated. The highest percentage does occur in the maximum caution class, but at 45% the disparity from the baseline of is not as large as for Asians.

Respondents who identified as White comprise 58% of the total sample. Their greatest discrepancy across the classes is to have the smallest percentage of those in the maximum caution class, 35%, and the largest percentage in the minimal concern group at 20% which is tied with Blacks.

The trends of race and gender considered separately are apparent when examined jointly. Fifty-nine percent of Asian females are in the class of maximum caution compared to 38% of White females. Fifty percent of Black females and 49% of Hispanic females are also in the maximum caution class.

Asian males membership in the maximum caution class is still high at 51%, but not as high as 59% for Asian females. Black and Hispanic male endorsement of maximum caution drops to 32% and 35%, respectively. White, male membership in this class declines all the way down to 29%, with 25% in the minimal concern group.

Age is related to the five latent classes (?² = 82.739, df = 20, p value = 0.000). As age increases, concern with precautionary measures regarding the pandemic tends to decrease. For the two youngest age groups up

to 25.5 years old, an average of 46% of students are in the maximum caution group, and only 15% and 15% of these students respectively express minimal concern. For the 59 students in the age group of 50 and above, only 20% were in the maximum caution group, and almost one third in the minimal concern group.

Cumulative GPA did not demonstrate a relationship with latent class (?² = 21.081, df = 16, p value = 0.175). Percentage of class membership is relatively constant across categories of GPA to within sampling error. For example, membership in the maximum concern group varies only from 38% to 41% across the five categories of GPA.

The length of time in the educational system is strongly associated with membership in the latent classes (?² = 54.430, df = 20, p value = 0.000). The longer a student, the less concerned with COVID precautions. For example, moving from freshmen to doctoral students displays monotonically diminishing group membership in maximal caution from 51% to 27%.

The instructional unit or academic discipline is defined as the most general area of study, individual colleges and schools within the university. The instructional unit also related to five identified latent classes (?² = 82.553, df = 32, p value = 0.000). Sample sizes varied from 121 for undeclared students to 1,051 for liberal arts and sciences.

Four different units, including undeclared, tied for the largest percentage of students, 42%, in the maximum caution class: also, liberal arts and sciences, public health, the arts.

Business, urban and public affairs, and engi- neering/computer science students in the maximum caution group were 34%, 30%, and 29%, respectively. Membership in the minimal concern group varied from a low of 16% for liberal arts and sciences and public health to highs of 24% for engineering and computer science and 26% for business.

Of course, liberal arts and sciences are composed of a wide variety of departments. Sample sizes for the different constitute departments are necessarily smaller, but no general difference between students in the physical sciences and social sciences was observed. For example, 58% of 36 sociology majors were in the maximum caution group, and so were 52% of 40 chemistry majors. The largest discrepancy was for mathematics majors, of which only 24% of 54 students were in the maximum caution group and 22% in the minimal concern group.

The analysis in this section attempts to understand how preference for online versus face-to-face (FSF) instruction impacts or was impacted by student membership in the various latent classes. To assess this, we present a scale for online preference and compare that to class membership.

Preference for F2F and online were separately assessed on 0 to 100 scales in increments of 10. Define a transformed variable, OnlF2F, to simultaneously scale both preferences as the difference between the two scores, online - F2F. A student with the strongest preference for Online would score 100 on online and 0 on F2F, resulting in an OnlF2F score of 100. A student with the strongest preference for F2F would score the maximum 100 on F2F and 0 on online, resulting in a minimum OnlF2F score of -100 that indicates the maximum preference for F2F. Students with similar preferences for both modalities would score approximately 0.

Overall, the majority preference is for F2F; 1,905 or 63.7% of students favored at least to some extent F2F teaching. The most favored response category is -100, a strong preference for F2F teaching, 469 or 15.7% of all respondents.

The decrease in preference scores does not monotonically decrease from -100 to 0. Instead, students who prefer F2F tend to have either a strong preference for F2F or a moderate preference. The average percentage of students who scored in the range of from -50 through -10 is only 32% compared to an average of only 17% from -90 through -60.

The second most frequent category is neutrality, a score of 0, indicated by 9.5% of the respondents. The third most frequent category is the strongest preference for online instruction, with 6% of all respondents scoring the maximum of 100. Moving from neutrality toward 100, the same pattern for the F2F scores is repeated, though with less intensity. There are more students with a mild preference for online than a strong but not the strongest preference. For students scoring from 10 through 40, the average percentage of students for each level is 12.1%. For students scoring from 50 through 90 the average percentage reduces to 9%.

Figure 16 demonstrates class membership across the span of F2F versus online preference scores. Class membership significantly varies across levels of online preference (?² = 223.341, df = 80, p value = 0.000).

For those students in the maximum caution class, about the same level of class membership is present at the two extreme positions and the neutral position, 38%, 36%, and 41%, respectively. The more dramatic differences occur for those in the minimal concern class.

Here only 14% of those students who prefer F2F are described by minimal concern, yet 43% of students in that class strongly prefer Online teaching. One explanation of this result is that what drives student preference for F2F is other issues such as flexibility of schedule and not concern with COVID per se. Those who wish to return to campus, however, are more concerned with COVID. Another explanation is that students who wish to pursue their studying entirely online are simply not worried about COVID on campus.

Similarly, those students who score 70 have by far the largest percentage of membership in the maximal concern group, 62%, followed by 53% for those who score 80. Compare this number with 36% membership for those who strongly endorse Online with a score of 100.

Latent class analysis of response patterns to the items on the survey indicate that precautionary measures in reaction to the COVID pandemic revealed five distinct classes of students that systematically varied in their level of concern for the implementation of these measures. The largest group of students, 37%, indicated maximum caution, preferring all cautionary possibilities. Another 19% of students preferred almost as much caution, though with no temperature screening before class meetings. Another group of 19% of students preferred measures that the university can control, though again with less emphasis on temperature screening. The smallest group of only 6% of students preferred only cautionary measures beyond control of the university per se, a low overall infection rate and an effective vaccine. Nineteen percent of students had minimal concern for any precautionary measures.

The composition of these groups differed according to race and gender. Females tended to prefer more precautionary measures than males. Asians as a group had the largest percentage of members in the maximum caution group, 56%, compared to only 35% of Whites in that group, with Black and Hispanic/Latino students intermediate with respective percentages of 43% and 45%. Asian women, accordingly, were the most risk-averse group, with 59% in the maximum concern group and only 13% in the minimal concern group. Only 29% of White men were in the most risk averse group, with 25% in the minimal concern group.

Regarding preference for F2F versus online teaching, the largest group of students tended to prefer F2F teaching, with 16% of students exhibiting a strong preference for F2F. However, the distribution of preference is diverse, with 6% of students indicating a strong preference for online. Preferences tended to cluster at the extremes of the distribution, or in the middle, the range of neutrality, with still a tendency toward the preference of F2F.

F2F versus online preference varied according to class membership, but more in terms of the minimal concern group than the maximum concern group. For an extreme scores of -100 (maximum F2F preference), a score of 0, and an extreme score of +100 (maximum online preference), respective class memberships of the minimal concern group increase from 14% to 23% to 43%. Yet the corresponding membership in the maximal concern group is relatively stable, 38%, 40%, and 36%.

The anomaly is that 65% of those who express a moderately strong endorsement of online (70), but not the strongest endorsement, express the most concern for COVID as reflected by group membership. One explanation for these results is that the level of COVID concern is motivating students to change their preference. Perhaps some students are so concerned about COVID that they prefer online but endorsed a moderately strong F2F position because of their concern for their education. Conversely, some students who prefer F2F might endorse a moderately strong online position because of their COVID concerns.

The COVID-19 pandemic has impacted schools and universities worldwide. The pandemic has forced total closure or near total closure of educational facilities at all levels. These school closures have impacted students, teachers, and their families, and are having far- reaching economic and societal consequences. As a result, it is imperative that schools reopen as soon as possible. This is top of mind for university leaders and administrators.

This article reports on a survey taken at one medium sized urban university about student attitudes toward returning to campus and the safety precautions universities can and should take to keep students safe. The response rate was relatively high (3,940 responses from a total of 21,593 students or 18.2%) The survey also links these preferences for COVID driven mitigations on campus to student preference for online versus face-to-face education. In addition, we include a detailed analysis of student sentiment analyzed by the student demographics of gender, race and ethnicity, age, academic discipline, and education level.

Overall, less than 15% of students feel safe to come back to campus immediately, and there is a larger percentage of students who would like to complete their educations remotely regardless of the mitigations put in place (22.0%). The majority of the students are somewhere in the middle. There are mitigations that the university can implement to help them feel safe to return to campus. These include mandatory mask wearing, social distancing in the classroom, temperature screening prior to entering the classroom and special COVID cleaning of the rooms between classes. There are significant differences in student feelings around safety based on gender, ethnicity/race, part-time versus full-time enrollment, student age, and instructional unit. These demographics are analyzed in this article. This study was part of a larger multiyear study into student preferences toward online education.

A latent class model was used which identified 5 distinct groupings of students requesting various degrees of safety precautions being put in place. Regarding preference for F2F versus online teaching, the largest group of students tended to prefer F2F teaching, with 16% of students exhibiting a strong preference for F2F. F2F versus online preferences tended to cluster at the extremes of the distribution, or in the middle, the range of neutrality, with still a tendency toward the preference of F2F. F2F versus online preference varied according to class membership, but more in terms of the minimal concern group than the maximum concern group.

Overall, more work needs to be done to better understand student concerns and to address their need to feel safe when they return to campus. Moreover, for planning purposes, university administrators also need to understand what mitigations students are willing to participate in so that the entire community of learners can be protected.

The purpose of this section is to provide information about the representativeness of sample respondents compared to the population. The survey distributed to all 21,593 registered students at Portland State University during October and November of 2020. Portland State is a midsize urban university located in Portland, Oregon. In this appendix we compare the percentage of respondent subgroups to those found within the population. The demographic categories analyzed in this article include age, race/ethnicity, gender/legal sex, cumulative GPA, student level and academic discipline.

As can be seen in Table A1, females (+9.5%) and high GPA students (+8%) were overrepresented in the sample as well as students ages 30-40 (+4%) and white students (+3%). Subgroups underrepresented in the sample compared to the population of all students were males (-9.5%), students with GPAs below 3.0 (-6.9%). Black, Hispanic, international, and unidentified students were each underrepresented by -1%.