Project Report

The Stability of Transport-Related Attitudes over Time: A

Case Study During COVID-19
Prepared for Teaching Old Models New Tricks (TOMNET) Transportation Center

By,

Laura Mirtich
Email: [email protected]

Matthew Wigginton Conway

Email: [email protected]

and others…

School of Geographical Sciences and Urban Planning

Arizona State University
975 S Myrtle Ave, Tempe, AZ 85287

June, 2022

1
TECHNICAL REPORT DOCUMENTATION PAGE
1. Report No. 2. Government Accession No. 3. Recipient's Catalog No.
N/A N/A N/A
4. Title and Subtitle 5. Report Date
The Stability of Transport-Related Attitudes over Time: A Case Study June 2022
During COVID-19 6. Performing Organization Code
N/A
7. Author(s) 8. Performing Organization Report No.
Laura Mirtich, https://orcid.org/0000-0002-0467-2934 N/A
Matthew W. Conway, https://orcid.org/0000-0002-1210-2982
Deborah Salon, https://orcid.org/0000-0002-2240-8408
Peter Kedron, https://orcid.org/0000-0002-1093-3416
Rishabh S. Chauhan, https://orcid.org/0000-0001-7188-557X
Sybil Derrible, https://orcid.org/0000-0002-2939-6016
Sara Khoeini, https://orcid.org/0000-0001-5394-6287
Abolfazl Kouros Mohammadian, https://orcid.org/0000-0003-3595-3664
Ehsan Rahimi,
Ram M. Pendyala, https://orcid.org/0000-0002-1552-9447
9. Performing Organization Name and Address 10. Work Unit No. (TRAIS)
School of Sustainable Engineering and the Built Environment N/A
Arizona State University 11. Contract or Grant No.
Tempe, AZ, 85281 69A3551747116
12. Sponsoring Agency Name and Address 13. Type of Report and Period Covered
U.S. Department of Transportation, Research Report
University Transportation Centers Program, (6/2020 – 12/2021)
1200 New Jersey Ave, SE, Washington, DC 20590 14. Sponsoring Agency Code
USDOT OST-R
15. Supplementary Notes
N/A
16. Abstract
The usefulness of attitudes in travel demand forecasting requires predictability. Since travel demand models aim to
simulate future populations, research would be impracticable if the characteristics of the populations were subject to
substantial unpredictable variation over time. We investigate the stability of individuals' attitudes using waves of the
COVID Future survey answered 3.5–9.5 months apart. Both individual attitudinal statements and factor-analyzed
attitudes demonstrate moderate stability. This stability is mostly consistent across different sub-populations, although
certain groups such as young people displayed lower levels of stability than the general population. Attitudes about home
environment and lifestyle were particularly stable, while those about pandemic-specific topics such as remote work or
disease risk were more unstable. We conclude that attitudes generally display stability, although the presence of significant
life disruptions likely produces temporary instability. We also demonstrate that the stability of attitudes can have an effect
on the stability of intended future actions.
17. Key Words 18. Distribution Statement
Attitude stability, Factor analysis, Panel analysis, Transport related No restrictions.
attitudes
19. Security Classif.(of this report) 20. Security Classif.(of this page) 21. No. of Pages 22. Price
Unclassified Unclassified 24 N/A
Form DOT F 1700.7 (8-72) Reproduction of completed page authorized
DISCLAIMER

The contents of this report reflect the views of the authors, who are responsible for the facts and
the accuracy of the information presented herein. This document is disseminated in the interest of
information exchange. The report is funded, partially or entirely, by a grant from the U.S.
Department of Transportation’s University Transportation Centers Program. However, the U.S.
Government assumes no liability for the contents or use thereof.

ACKNOWLEDGMENTS

This study was funded by a grant from A USDOT Tier 1 University Transportation Center,
supported by USDOT through the University Transportation Centers program. The authors would
like to thank the TOMNET, USDOT, and the US National Science Foundation for their support of
university-based research in transportation, and especially for the funding provided in support of
this project.

3
TABLE OF CONTENTS

TECHNICAL REPORT DOCUMENTATION PAGE .................................................................. 2

DISCLAIMER ................................................................................................................................ 3
ACKNOWLEDGMENTS .............................................................................................................. 3
EXECUTIVE SUMMARY ............................................................................................................ 5
INTRODUCTION .......................................................................................................................... 6
DATA ............................................................................................................................................. 7
STABILITY OF ATTITUDINAL STATEMENTS ....................................................................... 7
STABILITY OF ATTITUDINAL FACTORS ............................................................................. 11
ROBUSTNESS CHECK ACROSS SUB-SAMPLES .................................................................. 14
IMPACT OF ATTITUDINAL STABILITY ON BEHAVIORAL EXPECTATIONS ............... 18
CONCLUSION ............................................................................................................................. 18
REFERENCES ............................................................................................................................. 19
APPENDIX A: DEMOGRAPHICS OF SAMPLE ...................................................................... 22
APPENDIX B: WAVE 1 FACTOR ANALYSIS ........................................................................ 23

LIST OF TABLES
Table 1: Test-retest reliability of individual attitudinal statements ................................................ 9
Table 2: Summary of Waves 1 and 2 factor scores based on Wave 1 factor analysis .................. 14
Table 3: Attitudinal factor stability (ICC values) for different demographic groups ................... 15
Table 4: Attitudinal factor stability (ICC values) for different non-demographic groups ............ 16
Table 5: Attitudinal factor stability (ICC values) for groups with different pandemic experiences
....................................................................................................................................................... 17
Table A1: Demographics of respondents alongside Census data from 2018 1-year American
Community Survey ....................................................................................................................... 22
Table B1: Exploratory factor analysis (minimum residual method, varimax rotation) based on
Wave 1 data................................................................................................................................... 24

LIST OF FIGURES
Figure 1: Heatmaps of Wave 1 and Wave 2 responses to four attitudinal statements .................. 10
Figure 2: Histogram of changes in score between Wave 1 and Wave 2 for each factor from the
Wave 1 factor analysis .................................................................................................................. 12
Figure 3: Correlations between Wave 1 and Wave 2 scores for each factor from the Wave 1
factor analysis ............................................................................................................................... 13

4
EXECUTIVE SUMMARY
Attitudes and preferences are important for modeling travel behavior (Conway et al., 2020).
While the importance of attitudinal variables in predicting transport choices is mostly well-
established in the literature (de Abreu e Silva, 2014; Belgiawan et al., 2016; Kitamura et al.,
1997), some dissent comes from those who dispute the long-term stability (and therefore utility)
of at least certain attitudes (Jensen et al., 2013; Borriello & Rose, 2021). While some have found
stability in transport-related attitudes (van de Coevering et al., 2021; Willis & Lee, 1980), others
have reported instability over time frames ranging from one week to two years (Sunkanapalli et
al., 2000; Adams et al., 2013; Thøgersen, 2006).

We investigate the stability of individuals’ attitudes using waves of the COVID Future survey
answered 3.5–9.5 months apart. The survey included Likert-scaled indicators of travel-related
attitudes ranging from environmentalism to opinion on remote work. We compared both
individual attitudinal statement stability over time as well as factor-analyzed attitudinal stability
over time.

Both comparisons indicate moderate stability in attitudes. This stability is mostly consistent
across different sub-populations, although certain groups such as young people displayed lower
levels of stability than the general population. Attitudes about home environment and lifestyle
were particularly stable, while those about pandemic-specific topics such as remote work or
disease risk were more unstable. We conclude that attitudes generally display stability, although
the presence of significant life disruptions likely produces temporary instability. We also
demonstrate that the stability of attitudes can have an effect on the stability of intended future
actions.

The data collected in this project is publicly and permanently available on the ASU Dataverse
(https://dataverse.asu.edu/dataverse/covidfuture).

5
 Abstract
The usefulness of attitudes in travel demand forecasting requires predictability. Since
travel demand models aim to simulate future populations, research would be impracticable
if the characteristics of the populations were subject to substantial unpredictable variation over
time. We investigate the stability of individuals’ attitudes using waves of the COVID Future
survey answered 3.5–9.5 months apart. Both individual attitudinal statements and factor-
analyzed attitudes demonstrate moderate stability. This stability is mostly consistent across
different sub-populations, although certain groups such as young people displayed lower levels
of stability than the general population. Attitudes about home environment and lifestyle were
particularly stable, while those about pandemic-specific topics such as remote work or disease
risk were more unstable. We conclude that attitudes generally display stability, although
the presence of significant life disruptions likely produces temporary instability. We also
demonstrate that the stability of attitudes can have an effect on the stability of intended future
actions.

1 Introduction
Attitudes and preferences are important for modeling travel behavior (Conway et al., 2020). While
the importance of attitudinal variables in predicting transport choices is mostly well-established
in the literature (de Abreu e Silva, 2014; Belgiawan et al., 2016; Kitamura et al., 1997), some
dissent comes from those who dispute the long-term stability (and therefore utility) of at least
certain attitudes (Jensen et al., 2013; Borriello & Rose, 2021). While some have found stability
in transport-related attitudes (van de Coevering et al., 2021; Willis & Lee, 1980), others have
reported instability over time frames ranging from one week to two years (Sunkanapalli et al., 2000;
Adams et al., 2013; Thøgersen, 2006). Because of their correlation with relatively stable behaviors
(e.g., mode choice), we hypothesize that transport-related attitudes are themselves stable over time.
We also expect the stability of attitudes to be reflected in the outcomes they are hypothesized
to influence; that is, more stable attitudes should be held by people with more stable behavior.
We also hypothesize that stability will be lower for individual statements than for factors which
represent underlying attitudes indicated by multiple questions, since factors are less likely to reflect
idiosyncratic responses to individual questions.
The remainder of this article is organized as follows. Section 2 describes the data used in this
study, which was collected during the COVID-19 pandemic. Section 3 presents an analysis of
responses to single attitudinal statements across multiple survey waves. Section 4 describes our
methodology for factor analysis and summarizes the stability of attitudinal factors derived from the
statements appearing in section 3. In section 5, sub-samples of respondents based on demographics,
personal characteristics, survey experience, and pandemic experience are separately analyzed to
explore heterogeneity in attitudinal stability across the sample. Section 6 addresses the connection
between attitudinal stability and stability of expected behavior and Section 7 concludes with key
findings and suggestions for future research.

6
2 Data
Our data come from the first two waves of the COVID Future survey, which were administered
in June 2020–October 2020 and November 2020–May 2021 (Chauhan et al., 2021). COVID-19’s
impact on travel, shopping and dining habits, remote working, and learning were major topics
covered in the survey. Also included was a battery of attitudinal statements covering respondents’
perceptions and opinions related to these topics. Individual responses were recorded three and a
half to nine and a half months apart. About 50% of responses were recorded between four and
seven and a half months apart, with an average gap between responses of five and a half months.
To achieve a large, representative sample, we contacted respondents through survey organizations
(Data Axle and Qualtrics) using a quota-sampling method. The 2,673 respondents analyzed here
are fairly representative, though older, more educated, more likely to be female, and slightly higher
income than the population (Table A1). One potential source of sample bias is that our Wave 1
respondents were quota-sampled to be representative, but respondents had to return for the second
wave of the survey for this analysis. A higher likelihood of certain individuals to complete the
Wave 2 survey could contribute to a non-representative sample. All respondents were from the
United States, with 45 of 50 states and Washington, D.C. represented1 .

3 Stability of attitudinal statements

The survey included Likert-scaled indicators of travel-related attitudes ranging from environmentalism
to opinion on remote work. For each of the 26 attitudinal statements analyzed here, we recorded
the mean score for each of the two waves of the survey and the mean absolute difference in score,
with a one-point difference indicating a one-point shift on a five-point scale2 . We also reported
the percentage of respondents who answered a question identically in both waves of the survey
as well as the percentage of respondents whose change in answer between waves was zero or
one points. An intraclass correlation coefficient (ICC) is provided for each question. ICC is a
correlation coefficient used to measure the similarity of observations between paired waves of
data (Liljequist et al., 2019). As is recommended by Koo and Li (2016), a two-way mixed effect,
absolute agreement, single rater ICC is calculated. This form of ICC is appropriate since the data
involves a repeated sample of the same raters (”two-way”), cannot be perfectly generalized (”mixed
effects”), is not concerned with relative consistency (”absolute”), and deals with individual rather
than group consistency (”single rater”).
Travel-related attitudinal statements demonstrate moderate stability (Table 1). The mean absolute
difference in score was less than one for all statements. For most statements, only 50-60% of
respondents provided exactly the same answers in both waves, which could suggest that responses
are not stable. However, close 90% of respondents on average selected an answer that was the same
or adjacent on the Likert scale—for instance, transitioning from strongly disagree to somewhat
disagree. These small changes are unlikely to substantially affect modeling outcomes, particularly
once factor-analyzed.
Koo and Li (2016) suggest ICC cutoffs of 0.5, 0.75, and 0.9 to signify moderate, high, and
excellent stability, respectively. All statements have moderate stability, except one with high
1
The sample did not include any respondents from Alaska, Delaware, Hawaii, North Dakota, or Wyoming
2
Scores are coded from -2 to 2, with positive numbers indicating agreement

7
stability and one with low stability. The average ICC is 0.63 for an attitudinal question.
Figure 1 illustrates the stability of attitudinal statements by displaying heatmaps for the questions
with varying ICC values. Darker values indicate higher numbers of respondents associated with
a particular cell. All questions show a clustering of respondents along the bottom left-top right
diagonal, which contains the 5 cells associated with the same answer in both waves of the survey.

8
 Mean
Mean Percent Percent Intraclass
Wave 1 Wave 2 absolute exactly within correlation
difference matching 1 point coefficient

Shutting down businesses to prevent the spread of -0.57 -0.37 0.65 55% 86% 0.71
coronavirus is not worth the economic damage that
will result
I am concerned that friends or family members will 0.99 0.89 0.57 57% 90% 0.64
have a severe reaction to the coronavirus if they
catch it
Society is overreacting to the coronavirus -0.98 -0.96 0.47 68% 90% 0.76
If I catch the coronavirus, I am concerned that I will 0.69 0.53 0.61 56% 87% 0.69
have a severe reaction
Everyone should just stay home as much as possible 1.05 0.93 0.56 58% 90% 0.69
until the coronavirus has subsided
My friends and family expect me to stay at home until 0.32 0.24 0.81 44% 81% 0.54
the coronavirus subsides
I am committed to using a less polluting means of 0.21 0.18 0.62 51% 89% 0.66
transportation (e.g., walking, biking, and public
transit) as much as possible
I am committed to an environmentally-friendly 0.72 0.71 0.45 61% 95% 0.70
lifestyle
Sometimes I worry about the effects of airplane trips -0.01 0.00 0.65 52% 87% 0.69
on the environment
I dislike change 0.12 0.11 0.55 57% 90% 0.67
I enjoy spending time with the people I live with 1.22 1.22 0.39 68% 94% 0.65
The time I spend traveling to places provides a useful 0.66 0.66 0.67 49% 87% 0.42
transition between activities
Apartment living doesn’t provide enough privacy 0.56 0.58 0.70 50% 85% 0.56
Having shops and services within walking distance of 0.39 0.38 0.66 49% 88% 0.67
my home is important to me
I like to have a yard at home 1.25 1.25 0.39 68% 94% 0.72
Even if I do not end up buying anything, I still enjoy 0.33 0.31 0.68 52% 86% 0.66
going to stores and browsing
In-person shopping is usually a chore for me -0.14 -0.13 0.64 53% 88% 0.69
I enjoy shopping online 0.94 0.97 0.45 63% 94% 0.71
Online learning is a good alternative to high school- 0.16 0.07 0.84 43% 79% 0.53
and college-level classroom instruction
Video calling is a good alternative to in-person 0.78 0.78 0.67 50% 87% 0.50
business meetings
Video calling is a good alternative to visiting friends 0.25 0.17 0.81 46% 80% 0.55
and family
I am generally satisfied with my life 0.80 0.80 0.58 59% 89% 0.58
It is hard to get motivated to work away from the main -0.34 -0.34 0.72 51% 82% 0.51
office
I like working from home 0.52 0.47 0.58 59% 86% 0.64
I enjoy the social interaction found at a conventional 0.64 0.62 0.55 56% 91% 0.64
workplace
Learning how to use new technologies is often -0.09 -0.04 0.63 54% 87% 0.69
frustrating

Table 1: Test-retest reliability of individual attitudinal statements

9
Figure 1: Heatmaps of Wave 1 and Wave 2 responses to four attitudinal statements; Wave 1
responses are recorded along the x-axis and Wave 2 responses are recorded along the y-axis

10
4 Stability of attitudinal factors
Factor analysis is a strategy commonly used to identify attitudinal constructs by reducing a large
number of indicator statements to a smaller set of underlying factors (Conway et al., 2020). We
performed an exploratory factor analysis of the attitudinal statements to identify eight factors from
Wave 1 responses (Table B1). An eight-factor solution was chosen because it produces interpretable
factors with eigenvalues greater than 1.
Using the regression method, we estimated factor scores for both Wave 1 and Wave 2 responses.
Factor analysis is generally applied to standardized data (Grice, 2001). In order to make factor
scores directly comparable, we “standardized” the Wave 2 responses using the mean and standard
deviation of the Wave 1 data. We then scored both waves of responses using identical weights
generated from Wave 1 data. Once factor scores had been computed, we standardized them to
have mean 0 and standard deviation 1. Since the Wave 1 data was used for standardization, the
mean and standard deviation of Wave 2 factor scores were slightly different from 0 and 1, although
close. Because of these standardization and scoring techniques, Wave 1 and Wave 2 scores are
directly comparable and a 1-point change in a factor score can be interpreted as a one-standard
deviation change.
In line with the methodology of Adams et al. (2013), we calculated an ICC for the Wave 1 and
Wave 2 scores for each attitude. As with our analysis of individual questions, we calculated the
mean score difference and mean absolute score difference between waves, as well as the percentage
whose Wave 1 and Wave 2 scores were within one standard deviation of each other.
Our analysis shows that factor-analyzed attitudes are also moderately stable (Table 2). Changes
in factor scores are symmetrical around 0, showing no definite trend over time (Figure 2)3 . The
mean absolute changes in score are all fractions of a standard deviation. The changes in attitude that
did occur were fairly small in magnitude; 89–93% of respondents recorded a change in factor score
of less than one standard deviation. Using the guidelines recommended by Koo and Li (2016), all
attitudes display moderate stability.
An unexpected result is the lack of improved stability once attitudinal statements are factor
analyzed. Although the average ICC among factors (0.66) is slightly higher than that of questions
(0.63), the difference is not great. This contradicts previous research (Adams et al., 2013). The
higher stability of factors is generally found because idiosyncratic responses to certain questions
reduce question-level stability more dramatically than they reduce factor-level stability. Such
idiosyncratic responses may have been uncommon in this data set.
As a visual presentation of factor-level stability, respondents’ Wave 1 and Wave 2 scores for a
given factor were plotted (Figure 3). These scatter plots show clustering around the identity line,
indicating that most respondents have Wave 2 scores very close to their Wave 1 scores4 .
3
The unusual distribution of the ”Pro-In-Person-Shopping” histogram is a result of this factor score being
determined almost entirely by a single question (see Table B1)
4
The unusual cloud of points on the ”Pro-In-Person-Shopping” scatter plot is a result of this factor score being
determined almost entirely by a single question (see Table B1)

11
Figure 2: Histogram of changes in score between Wave 1 and Wave 2 for each factor from the
Wave 1 factor analysis 12
Figure 3: Correlations between Wave 1 and Wave 2 scores for each factor from the Wave 1 factor
analysis 13
 Mean change Mean absolute Percent
ICC
in score change in score within 1 SD
Anti-lockdown 0.04 0.57 91% 0.68
Environmentalist 0.03 0.56 92% 0.71
city lover
Pro- -0.03 0.67 89% 0.57
videoconferencing
Disease concerned -0.08 0.67 91% 0.58
Pro-in-person- -0.00 0.57 92% 0.66
shopping
Anti-working from 0.00 0.63 90% 0.64
home
Social optimistic -0.00 0.60 92% 0.70
Anti-apartment 0.03 0.56 93% 0.74

Table 2: Summary of Waves 1 and 2 factor scores based on Wave 1 factor analysis

5 Robustness check across sub-samples

We also analyzed the attitudinal stability of specific sub-populations of respondents. First, we
grouped respondents by demographics and personal characteristics such as age, gender, education
status, industry of employment, and household size. For this analysis, we considered high income
respondents to be those making over $100,000 annually and low income respondents to be those
making under $50,000 annually. Large households were those containing more than three people.
Finally, the cutoff for living an an urban environment was a zip code-level housing density of at
854.4 housing units per square kilometer and the cutoff for a rural environment was less than 39.4
housing units per square kilometer Kolko, 2015.
Next, we created groups based on when respondents took the survey, the gap between Wave 1
and Wave 2 responses, and the manner through which the respondent was recruited. A long gap
between survey waves was at least 135 days and an early respondent took the Wave 1 survey by
the end of July 2020.
Finally, we categorized respondents by whether they had experienced certain events such as
transitioning to remote work, contracting COVID-19, and living under various lockdown measures.
For each group, we replicated the factor-level analysis outlined above and provided an ICC for each
factor based on only members of that particular group. This served the dual purpose of acting as
a robustness check and allowing for more detailed conclusions to be drawn about attitude stability
across the population.
Two of the most important demographic characteristics that appear to impact stability are age
and education level. Respondents younger than 35 have lower stability, a result consistent with
other findings from the literature (Alwin & Krosnick, 1991; Visser & Krosnick, 1998). Likely
because most are under age 35, students in the sample have less stable attitudes than non-students.
In addition to age, higher levels of educational attainment are associated with higher attitudinal
stability. Workers who describe themselves as being employed in a professional field report their
own attitudes more consistently than non-workers or those employed in other fields. Similar to the

14
case of students and young people, the attitudinal stability of professionals likely results from most
members of this group holding at least an undergraduate degree.

Age Bach. No bach. High

Men Women Age<35 Age≥65
Factor 35-64 degree degree income
(n=925) (n=1732) (n=340) (n=926)
(n=1365) (n=1465) (n=1207) (n=777)
Anti-lockdown 0.69 0.68 0.62 0.68 0.70 0.70 0.66 0.67
Environmentalist
city lover 0.71 0.71 0.63 0.73 0.72 0.73 0.69 0.70
Pro-
videoconferencing 0.58 0.57 0.58 0.59 0.55 0.61 0.52 0.60
Disease concerned 0.53 0.61 0.57 0.60 0.58 0.54 0.63 0.56
Pro-in person
shopping 0.67 0.64 0.53 0.65 0.68 0.68 0.62 0.70
Anti-working
from home 0.63 0.65 0.65 0.64 0.61 0.69 0.58 0.69
Social optimistic 0.69 0.70 0.66 0.72 0.70 0.71 0.68 0.67
Anti-apartment 0.74 0.74 0.68 0.73 0.77 0.73 0.75 0.68
Mid Low Small Large Sub- Essential
Urban Rural
Factor income income household household urban worker
(n=847) (n=193)
(n=883) (n=937) (n=1719) (n=946) (n=1618) (n=573)
Anti-lockdown 0.71 0.67 0.70 0.65 0.68 0.69 0.69 0.64
Environmentalist
city lover 0.70 0.73 0.73 0.68 0.70 0.70 0.71 0.66
Pro-
videoconferencing 0.60 0.53 0.56 0.60 0.61 0.55 0.58 0.49
Disease concerned 0.62 0.56 0.58 0.59 0.62 0.56 0.60 0.53
Pro-in person
shopping 0.69 0.61 0.68 0.61 0.63 0.67 0.68 0.59
Anti-working
from home 0.62 0.62 0.64 0.64 0.64 0.66 0.54 0.57
Social optimistic 0.66 0.71 0.72 0.66 0.72 0.70 0.68 0.67
Anti-apartment 0.76 0.75 0.76 0.68 0.76 0.70 0.69 0.67
Profes- Admin- Other Non- Non-
Student All
Factor sional istrator worker worker student
(n=178) (n=2673)
(n=558) (n=112) (n=136) (n=1293) (n=2495)
Anti-lockdown 0.71 0.60 0.68 0.70 0.57 0.69 0.68
Environmentalist
city lover 0.73 0.59 0.73 0.73 0.62 0.72 0.71
Pro-
videoconferencing 0.71 0.66 0.53 0.54 0.55 0.58 0.57
Disease concerned 0.59 0.64 0.63 0.59 0.56 0.58 0.58
Pro-in person
shopping 0.75 0.67 0.71 0.64 0.51 0.67 0.66
Anti-working
from home 0.73 0.74 0.68 0.57 0.60 0.65 0.64
Social optimistic 0.73 0.67 0.69 0.70 0.62 0.71 0.70
Anti-apartment 0.75 0.61 0.75 0.77 0.58 0.75 0.75

Table 3: Attitudinal factor stability (ICC values) for different demographic groups

In addition to personal characteristics, respondents were also grouped by how and when they
took the COVID Future survey (Table 4). Both waves of data collection took place over fairly
broad time periods. As a result, respondents varied widely in how long they waited between

15
taking the first and second waves. Unsurprisingly, respondents with a short gap reported their
attitudes more consistently than those who were surveyed over a longer time period. However, the
difference between these two groups for many questions is not large, which is an encouraging sign
that attitudinal stability does not deteriorate rapidly as the time scale for repeated measurements
moves from the short-term to the medium-term.
Attitude stability also varied by when respondents were surveyed. Those who took Wave 1
early in the pandemic provided more stable responses than those who were initially surveyed later.
Major life events or disruptions have been found by some to cause attitudinal change (Hatemi,
2013; Janke & Handy, 2019), and the COVID-19 pandemic will undoubtedly cause permanent
attitudinal changes for many. The timing and mechanisms by which pandemic experience alters
attitude are likely dependent on an individual’s preexisting opinions, the severity of the pandemic
in their region (and the timing of this severity), governmental response, and other context-specific
characteristics. While regional variation existed, the severity of the COVID-19 pandemic in the
United States generally worsened toward the end of 2020 and peaked in early 2021. It is possible
that people surveyed during this period of severe outbreak were experiencing a greater magnitude
of disruption in their lives and therefore had less stable attitudes.
Respondents were contacted by a variety of means for the COVID Future survey. A Qualtrics
opinion panel was used in addition to direct emails via Data Axle from both the University of Illinois
at Chicago (UIC) and Arizona State University (ASU). Respondents contacted by UIC via email
had the lowest average attitudinal stability, but this is largely due to their extremely low stability
on the Disease concerned factor; the three groups showed similar stability for all other attitudes

Long Short ASU UIC

Early Late Qualtrics All
Factor gap gap group group
(n=1924) (n=749) (n=1776) (n=2673)
(n=1447) (n=1198) (n=710) (n=187)
Anti-lockdown 0.67 0.71 0.69 0.67 0.69 0.66 0.66 0.68
Environmentalist
city lover 0.68 0.74 0.73 0.67 0.72 0.69 0.70 0.71
Pro-
videoconferencing 0.58 0.57 0.56 0.61 0.56 0.59 0.52 0.57
Disease concerned 0.55 0.62 0.60 0.53 0.63 0.52 0.35 0.58
Pro-in person
shopping 0.65 0.66 0.66 0.66 0.65 0.67 0.63 0.66
Anti-working
from home 0.63 0.66 0.66 0.60 0.62 0.71 0.59 0.64
Social optimistic 0.68 0.73 0.73 0.64 0.69 0.69 0.70 0.70
Anti-apartment 0.71 0.76 0.74 0.73 0.75 0.70 0.73 0.74

Table 4: Attitudinal factor stability (ICC values) for different non-demographic groups

A final robustness check examined whether pandemic experience affected attitudinal stability
(Table 5). The health risk posed by COVID-19 and a transition to remote work have been two
of the pandemic’s most wide-reaching impacts. However, personal experience with either did not
have an impact on attitude stability. Those who believed themselves or a household member had
COVID-19 did not exhibit markedly different stability from those who did not believe anyone in
their household had the disease. Similarly, attitude stability did not differ noticeably across those
who had experience working from home before pandemic, those who began working from home
for the first time during the pandemic, and those who were in-person workers both before and

16
during the pandemic. On the other hand, our survey indicates that experiences such as lockdowns
that affected groups of people more evenly did have an impact on attitudinal stability. People who
experienced closures of schools or businesses reported more consistent attitudes than those who did
not experience these, perhaps because the severity of COVID-19 was mitigated by such policies.
No similar effects were found for those who experienced mask mandates and social distancing or
stay-at-home orders.
Moving has been a unique experience during the pandemic, especially among those who have
chosen to move as a direct result of a transition to online work. Movers reported greater attitudinal
change between waves than non-movers. This is unsurprising, given that movers would have
experienced a life disruption of greater magnitude than non-movers.

Had No Previous New No No

Masks Distance
Factor COVID-19 COVID-19 WFH WFH WFH masks
(n=2500) (n=2484)
(n=390) (n=2283) (n=395) (n=344) (n=474) (n=173)
Anti-lockdown 0.65 0.69 0.65 0.69 0.68 0.68 0.68 0.68
Environmentalist
city lover 0.67 0.72 0.67 0.69 0.69 0.71 0.76 0.71
Pro-
videoconferencing 0.55 0.58 0.63 0.62 0.58 0.58 0.48 0.58
Disease concerned 0.57 0.59 0.57 0.58 0.56 0.58 0.56 0.58
Pro-in person
shopping 0.68 0.65 0.68 0.70 0.65 0.66 0.58 0.66
Anti-working
from home 0.67 0.64 0.64 0.69 0.59 0.65 0.53 0.64
Social optimistic 0.68 0.71 0.71 0.67 0.71 0.70 0.75 0.70
Anti-apartment 0.74 0.74 0.70 0.73 0.72 0.73 0.80 0.74
No No Non
Closures Movers All
Factor distance closures movers
(n=2493) (n=150) (n=2673)
(n=189) (n=180) (n=2523)
Anti-lockdown 0.78 0.69 0.56 0.63 0.69 0.68
Environmentalist
city lover 0.73 0.71 0.70 0.66 0.71 0.71
Pro-
videoconferencing 0.46 0.58 0.48 0.63 0.57 0.57
Disease concerned 0.67 0.58 0.60 0.46 0.59 0.58
Pro-in person
shopping 0.63 0.67 0.48 0.62 0.66 0.66
Anti-working
from home 0.62 0.65 0.54 0.57 0.65 0.64
Social optimistic 0.77 0.71 0.64 0.59 0.71 0.70
Anti-apartment 0.71 0.74 0.74 0.57 0.75 0.74

Table 5: Attitudinal factor stability (ICC values) for groups with different pandemic experiences

An in-depth robustness check reveals that some key groups of respondents experienced unusually
stable or unstable attitudes during the COVID-19 pandemic. Despite this heterogeneity, all 42 sub-
samples of respondents show fairly consistent levels of stability. Attitudinal factors with stability
outside the “moderate” range (ICC greater than 0.75 or less than 0.5) were rare.

17
6 Impact of attitudinal stability on behavioral expectations
Finally, we explored the relationship between attitude stability and behavioral stability. Our hypothesis
is that attitudes should be fairly stable since they are often found to be important predictors of
stable behaviors such as mode choice. Evidence for this relationship would be found in the form
of a correlation between attitudinal stability and behavioral stability. The COVID Future survey
gathered information from respondents on their expectations of future behaviors such as working
from home, using various daily travel modes, flying, dining in a restaurant, taking online classes,
and more. We selected attitudes and behaviors with a plausible relationship to each other (for
example, the Environmentalist city lover attitude and flying for personal trips) and performed an
analysis of variance (ANOVA) to determine the relationship between attitude stability and stability
of expected behavior. Respondents were grouped based on how different their Wave 1 and Wave
2 expectations of behavior were (an ordinal scale was used to report expected future behavior; for
personal airplane trips, respondents could report ”much less [than before COVID-19]”, ”somewhat
less”, ”about the same”, ”somewhat more”, or ”much more” – other behavioral outcomes used
this scale or a similar one). For example, a respondent might be classified as reporting identical
expectations in both waves of the survey, shifting one rank along the ordinal scale, shifting two
ranks along the ordinal scale, and so on. Then, an ANOVA was used to determine if absolute
changes in factor score between waves differed among groups.
The stability of the Environmentalist city lover, Pro-videoconferencing, Disease concerned, and
Pro-in-person shopping factors were all found to be related to the stability of one or more expected
future behaviors. Variable scores for the Environmentalist city lover factor were found to be more
common among those whose expectations for future pedestrian travel and future personal airplane
travel changed between waves. Stability in expectations of personal airplane travel were also
related to stability of the Disease concerned attitude. The stability of the Pro-videoconferencing
attitude was correlated with the stability of students’ desire to take online classes in the future.
Finally, a stable Pro-in-person shopping attitude was associated with stable expectations about
ordering groceries online for delivery. To the extent that the stated preference data of the COVID
Future survey will be reflective of actual future behavior, attitude stability and behavioral stability
exhibit a relationship in many cases.
Notably, some expected associations between stable attitudes and stable behavioral expectations
were not found in our data. One notable example is that the variability of the Anti-working-from-
home attitude has no impact on how constant respondents’ expectations about actually working
from home were. Similarly, an unstable level of Disease concern had no association with unstable
expectations about future use of public transit. Along with other insignificant tested relationships,
these examples show that the association between attitudinal stability and expected behavioral
stability is not universal.

7 Conclusion
We conclude that attitudes are moderately stable over a period of three and a half to nine and a half
months. Despite our hypothesis to the contrary, we find that the stability of individual attitudinal
statements is comparable to that of factor-analyzed attitudes. We also find multiple instances of
attitudinal stability being predictive of expected behavioral stability. This supports the hypothesis

18
that attitudinal stability can, to an extent, be inferred from the stability of the outcomes which it
predicts.
Our analyses also suggest that there is varied stability between different attitudes. ‘Pro-videoconferencing,
Disease concerned, and Anti-working from home are the least stable attitudes across the general
sample as well as many demographic subgroups, possibly because the pandemic has caused many
people to experience remote work, virtual communication technologies, and public health concerns
in a way that was new to them. Consistent with findings from the field of psychology, the increased
familiarity with these ways of life may induce positive or negative attitudinal changes (Moreland
& Zajonc, 1982; Norton & Frost, 2007). On the other hand, the most stable attitudes in the survey
were those which likely were affected least by the COVID-19 pandemic. The Anti-apartment,
Environmentalist city lover, and Social-optimistic attitudes tended to change less between survey
waves than explicitly pandemic-related attitudes.
Nevertheless, the pervasive effects of COVID-19 likely influenced the stability of even these
tangentially related attitudes – urban life and socialization have certainly undergone changes during
COVID-19. Psychological research suggests that such major life disruptions are associated with
attitude change (Hatemi, 2013), and this relationship has been found for travel-related attitudes
in particular (Janke & Handy, 2019). COVID-19 constitutes a major life event for many, so the
stability observed here could be unusually low. In fact, some other reviews of the stability of
transport-related attitudes find higher question-level ICCs of 0.66 to 0.77 (Molina-Garcia et al.,
2010), which suggests greater reliability of attitudinal indicators under more typical circumstances.
Even in this atypical time, however, attitudes are moderately stable.
There remain major gaps in the literature that require future research. This paper presented an
exploratory analysis of the relationship between stable attitudes and stable behavioral expectations.
This topic, particularly in the context of travel-related attitudes and transport choices, requires
further research. Additionally, a detailed discussion of how geography affected attitudes during
COVID-19 would be a valuable addition to the literature. People’s attitudes – and also their
attitudinal stability – is likely dependent in part on the severity of COVID-19 in their area, as
well as the responses of their government and peers. Further investigation into this link will be
important to understanding the long-term societal changes that COVID-19 is certain to cause.

References
Adams, E., Goodman, A., Sahlqvist, S., Bull, F., & Ogilvie, D. (2013). Correlates of walking and
cycling for transport and recreation: Factor structure, reliability and behavioural associations
of the perceptions of the environment in the neighbourhood scale (PENS). International
Journal of Behavioral Nutrition and Physical Activity, 10. https://doi.org/10.1186/1479-
5868-10-87
Alwin, D., & Krosnick, J. (1991). Aging, Cohorts, and the Stability of Sociopolitical Orientations
Over the Life Span. American Journal of Sociology, 97(1), 169–195. https://doi.org/10.
1086/229744
Belgiawan, P., Schmöcker, J., & Fujii, S. (2016). Understanding car ownership motivations among
Indonesian students. International Journal of Sustainable Transportation, 10, 295–307.
https://doi.org/10.1080/15568318.2014.921846

19
Borriello, A., & Rose, J. (2021). Global versus localised attitudinal responses in discrete choice.
Transportation, 48, 131–165. https://doi.org/10.1007/s11116-019-10045-3
Chauhan, R., Conway, M. W., Capasso da Silva, D., Salon, D., Shamshiripour, A., Rahimi, E.,
Khoeni, S., Mohammadian, A., Derrible, S., & Pendyala, R. (2021). A database of travel-
related behaviors and attitudes before, during, and after COVID-19 in the United States
[revision submitted, preprint available at https : / / arxiv . org / abs / 2103 . 16012]. Scientific
Data.
Conway, M. W., Mirtich, L., Salon, D., Harness, N., Ross, A., & Hong, S. (2020). How important
are attitudes in travel behavior models? [Presented at Bridging Transportation Researchers,
presentation available at https://bridgingtransport.org/blog/archive-of-jan-2020-meeting/].
de Abreu e Silva, J. (2014). Spatial self-selection in land-use travel behavior interactions: Accounting
simultaneously for attitudes and socioeconomic characteristics. Journal of Transport and
Land Use, 7, 63–84. https://doi.org/10.5198/jtlu.v7i2.696
Grice, J. W. (2001). Computing and evaluating factor scores. Psychological methods, 6(4), 430–
450. Retrieved January 8, 2021, from http://search.proquest.com/docview/72384053?pq-
origsite=summon
Hatemi, P. (2013). The Influence of Major Life Events on Economic Attitudes in a World of Gene-
Environment Interplay. American Journal of Political Science, 57(4), 987–1007. https://
doi.org/10.1111/ajps.12037
Janke, J., & Handy, S. (2019). How life course events trigger changes in bicycling attitudes and
behavior: Insights into causality. Travel Behaviour and Society, 16, 31–41. https://doi.org/
10.1016/j.tbs.2019.03.004
Jensen, A., Cherchi, E., & Mabit, S. (2013). On the stability of preferences and attitudes before
and after experiencing an electric vehicle. Transportation Research Part D: Transport and
Environment, 25, 24–32. https://doi.org/10.1016/j.trd.2013.07.006
Kitamura, R., Mokhtarian, P., & Laidet, L. (1997). A micro-analysis of land use and travel in
five neighborhoods in the San Francisco Bay Area. Transportation, 24, 125–158. https :
//doi.org/10.1023/A:1017959825565
Kolko, J. (2015). Urban Headwinds, Suburban Tailwinds. Trulia Research.
Koo, T., & Li, M. (2016). A guideline of selecting and reporting intraclass correlation coefficients
for reliability research. Journal of Chiropractic Medicine, 15(2), 155–163. https://doi.org/
10.1016/j.jcm.2016.02.012
Liljequist, D., Elfving, B., & Roaldsen, K. (2019). Intraclass correlation – A discussion and demonstration
of basic features. PLOS One, 14, 131–165. https://doi.org/10.1371/journal.pone.0219854
Molina-Garcia, J., Castillo, I., & Sallis, J. (2010). Psychosocial and environmental correlates of
active commuting for university students. Preventive Medicine, 51, 136–138. https://doi.
org/10.1016/j.ypmed.2010.05.009
Moreland, R., & Zajonc, R. (1982). Exposure effects in person perception: Familiarity, similarity,
and attraction. Journal of Experimental Social Psychology, 18(5), 395–415. https://doi.org/
10.1016/0022-1031(82)90062-2
Norton, M., & Frost, J. (2007). Less is more: The lure of ambiguity, or why familiarity breeds
contempt. Journal of Personality and Social Psychology, 92(1), 97–105. https://doi.org/10.
1037/0022-3514.92.1.97

20
Sunkanapalli, S., Pendyala, R., & Kuppam, A. (2000). Dynamic Analysis of Traveler Attitudes
and Perceptions Using Panel Data. Transportation Research Record, 1718, 52–60. https:
//doi.org/10.3141/1718-07
Thøgersen, J. (2006). Understanding repetitive travel mode choices in a stable context: A panel
study approach. Transportation Research Part A: Policy and Practice, 40(8), 621–638.
https://doi.org/10.1016/j.tra.2005.11.004
van de Coevering, P., Maat, K., & van Wee, B. (2021). Causes and effects between attitudes,
the built environment and car kilometres: A longitudinal analysis. Journal of Transport
Geography, 91, 102982. https://doi.org/10.1016/j.jtrangeo.2021.102982
Visser, P. S., & Krosnick, J. A. (1998). Development of attitude strength over the life cycle: Surge
and decline. Journal of Personality and Social Psychology, 75(6), 1389–1410. https://doi.
org/10.1037/0022-3514.75.6.1389
Willis, C., & Lee, J. W. (1980). Time Stability of Attitudes Toward Transit Use in the Orlando,
Florida, Urbanized Area [Presented at the 59th Annual Meeting of the Transportation Research
Board, available in Transportation Research Record, issue 761, pages 56-58.].

21
A Demographics of Sample
Table A1 shows the demographics of the respondents in our sample, who were older, more educated,
and slightly higher income than the overall population. The sample also heavily over-represented
women.

Survey Census
Education No school completed – 1.5%
respondents 25 years Some grade/high school 0.7% 10.2%
old or older Completed high school or GED 12.7% 26.9%
(n=2672) Some college or technical school 31.7% 28.9%
Bachelor’s degree(s) or some graduate school 33.0% 20.0%
Completed graduate degree(s) 21.8% 12.6%
Income Less than $10,000 3.2% 6.3%
(n=2597) $10,000 to $14,999 3.7% 4.3%
$15,000 to $24,999 7.6% 9.0%
$25,000 to $34,999 9.4% 8.9%
$35,000 to $49,999 12.1% 12.4%
$50,000 to $74,999 18.9% 17.4%
$75,000 to $99,999 15.1% 12.6%
$100,000 to $124,999 11.4% 9.2%
$125,000 to $149,999 5.8% 5.8%
$150,000 to $199,999 6.2% 6.6%
$200,000 or more 6.5% 7.6%
Age 18–24 3.2% 12.1%
(n=2631) 25–34 9.8% 17.9%
35–49 20.6% 24.5%
50–64 31.3% 24.9%
65 and over 35.2% 20.7%
Gender Female 64.9% 50.8%
(n=2668) Male 34.7% 49.2%
Other 0.4% –
Sample size 2673

Table A1: Demographics of respondents alongside Census data from 2018 1-year American
Community Survey

22
B Wave 1 factor analysis
Presented here is the factor analysis that reduces 26 attitudinal statements into 8 underlying factors.
Statements from Table 1 that do not appear here did not have a loading greater than ±0.3 for any
factor.

23
 Environ- Pro- Pro-In- Anti-
Anti- mentalist Video- Disease Person Working Social Anti-
Lockdown Concerned Optimistic Apartment
City Lover Conferencing Shopping From Home
Shutting down businesses to prevent the spread 0.781 – – – – – – –
of coronavirus is not worth the economic
damage that will result
I am concerned that friends or family members – – – 0.805 – – – –
will have a severe reaction to the coronavirus
if they catch it
Society is overreacting to the coronavirus 0.803 – – – – – – –
If I catch the coronavirus, I am concerned that I – – – 0.684 – – – –
will have a severe reaction
Everyone should just stay home as much as -0.556 – – – – – – –
possible until the coronavirus has subsided
I am committed to using a less polluting means – 0.863 – – – – – –
of transportation (e.g., walking, biking, and
public transit) as much as possible
I am committed to an environmentally-friendly – 0.687 – – – – – –
lifestyle
Sometimes I worry about the effects of airplane – 0.580 – – – – – –
trips on the environment
I dislike change – – – – – – -0.348 –
24

I enjoy spending time with the people I live with – – – – – – 0.345 0.317
Apartment living doesn’t provide enough – – – – – – – 0.467
privacy
Having shops and services within walking – 0.393 – – – – – –
distance of my home is important to me
I like to have a yard at home – – – – – – – 0.541
Even if I do not end up buying anything, I still – – – – 0.959 – – –
enjoy going to stores and browsing
In-person shopping is usually a chore for me – – – – -0.520 – – –
Online learning is a good alternative to – – 0.600 – – – – –
high school- and college-level classroom
instruction
Video calling is a good alternative to in-person – – 0.728 – – – – –
business meetings
Video calling is a good alternative to visiting – – 0.692 – – – – –
friends and family
I am generally satisfied with my life – – – – – – 0.326 –
It is hard to get motivated to work away from the – – – – – 0.666 – –
main office
I like working from home – – – – – -0.415 – –
I enjoy the social interaction found at a – – – – – – 0.431 –
conventional workplace
Learning how to use new technologies is often – – – – – 0.419 – –
frustrating

Table B1: Exploratory factor analysis (minimum residual method, varimax rotation) based on Wave 1 data. Only loadings greater than
±0.3 are shown.