1. Questions

COVID-19 has greatly impacted travel around the globe. Passenger-trips by all commercial airlines fell 50% compared to 2019 flights (OAG Aviation Worldwide 2021). Public transit fell sharply everywhere (Beck and Hensher 2020) largely because people hesitated to use shared modes (Dean and Zuniga-Garcia 2022; Loa et al. 2022). Although the pandemic’s impacts on daily travel behavior have been extensively explored (Abdullah et al. 2020; De Vos 2020; Habib et al. 2021), studies about impacts on long-distance (LD) travel are lacking. To examine such impact on Americans’ LD travel (with a focus on the state of Texas), a survey was conducted in March 2021, asking over 1,000 Americans about their business and non-business trip counts in 2019 and 2020. In this paper, LD travel is defined as at least 75 miles one-way (as also used in National Household Travel Survey 2016/2017 for Texas add-on questions), although many researchers choose 50 or 100 miles (Dowds et al. 2018).

2. Methods

The web-based survey, using Qualtrics software, had a three-week collection process. After data cleaning, the final response pool includes 451 (45%) Texans and 553 (55%) non-Texan Americans. A population-weighting procedure matched the demographic distributions of age, region, and gender based on the last 5 years of American Community Survey data (US Census Bureau 2021a), leveraging the iterative proportional fitting method (Roth, Dai, and DeMatteis 2017).

To capture the factors that impact LD trip-making frequency during the pandemic, this paper first presents the shares of LD trip-making occurrence in 2019 and 2020. The year 2019 has no pandemic impacts, in contrast to 2020. Negative binomial models were estimated to predict the numbers of LD business and non-business trips in 2019 and 2020, followed by a sensitivity analysis on all independent variables.

3. Findings

Demographic distributions of LD non-business and business trip counts during 2019 and 2020 are shown in Table 1. The effect of COVID-19 on LD trips was felt equally across non-business and business trips in general. Business trips accounted for about one-third of the total LD trips in 2019 and 2020. For non-business trips, females and males made about the same number of LD trips (with 1 to 2% more made by females), but males had more business trips, about twice of the amount made by females before the pandemic. People aged 25 to 34 years contributed the most to LD trip-making, with over a quarter for non-business purposes and over half for business purposes. While early-career individuals have a larger share of LD business trips, the share of LD non-business trips by age group is nearly uniform, perhaps because all age groups travel for vacations and to see family during holidays. People aged 25 to 54 made 81% of the LD business trips prior to the pandemic, and the number climbed up to 88% during the pandemic, which could be the possible delegation of work to lower-risk employees. In terms of travelers living in different census regions, the Western population made more business LD trips, on average, compared to other regions before pandemic, while Northeastern residents made more non-business trips. The split remained similar during the pandemic in 2020, but LD trip makers in the Southern US were less impacted by COVID-19 compared to the other three census regions, possibly due to the looser mask-wearing restrictions.

Table 1.Demographic distribution of long-distance trip occurrence in 2019 and 2020 in the US.
Calendar year 2019 2020
Trip purpose Non-Business Business Non-Business Business
66% 34% 67% 33%
Gender
Female 52% 34% 51% 38%
Male 48% 66% 49% 62%
Age
18 to 24 years 5% 5% 6% 7%
25 to 34 years 26% 52% 27% 55%
35 to 44 years 15% 12% 17% 13%
45 to 54 years 15% 19% 18% 20%
55 to 64 years 19% 9% 20% 4%
65 or more years 20% 3% 11% 1%
Region *2019 Population (%)
Northeast 17% 28% 14% 26% 15%
Midwest 21% 12% 5% 11% 4%
South 38% 31% 35% 34% 41%
West 24% 30% 47% 29% 41%

Note: 2019 population statistics were obtained from US Census Bureau (2021b).

A negative binomial (NB) regression model was conducted to estimate Americans’ LD business and non-business trip-making frequency in 2019 and 2020 (Table 2). Practical significance of the NB model estimates is shown in Figure 1, obtained through a sensitivity analysis. The practical significance is calculated as the average change (in percentage) in the estimated predictor by increasing the continuous variables one standard deviation or shifting categorical variables from 0 to 1.

Table 2.Parameter estimates in annual business and non-business trip counts prediction in 2019 and 2020 (using population-weighted negative binomial count model)
Business trip frequency in 2019 Business trip frequency in 2020
Estimate Std. Error P-value Estimate Std. Error P-value
(Intercept) -1.528 0.327 0.000 -2.385 0.444 0.000
Female -0.147 0.194 0.449 -0.067 0.261 0.796
Age 18-24 -1.373 0.317 0.000 -0.722 0.411 0.079
Age 65+ -0.702 0.335 0.036 -1.150 0.459 0.012
Resident in Northeastern US -0.025 0.268 0.927 -0.021 0.352 0.952
Resident in Midwestern US -1.228 0.279 0.000 -1.553 0.372 0.000
Resident in Western US 1.264 0.234 0.000 0.347 0.319 0.276
Education high school or higher 0.660 0.265 0.013 0.533 0.355 0.134
Income (in $10k) 0.461 0.207 0.026 0.462 0.277 0.095
Number of workers 0.375 0.094 0.000 0.458 0.126 0.000
Number of children 0.369 0.109 0.001 0.126 0.146 0.390
Full-time employed 1.367 0.271 0.000 1.655 0.360 0.000
Part-time employed 1.448 0.334 0.000 1.596 0.447 0.000
Married -0.460 0.233 0.048 0.398 0.317 0.210
Divorced -1.863 0.408 0.000 -1.078 0.530 0.042
No. of observations 1,004 1,004
Dispersion Parameter (ρ): 0.159 0.087
McFadden's R2: 0.332 0.304
Likelihood ratio test (χ2) 212 128
Prob > χ2 0.000 0.000
2 x log-likelihood -2,261 -1,608
Non-business trip frequency in 2019 Non-business trip frequency in 2020
Estimate Std. Error P-value Estimate Std. Error P-value
(Intercept) 1.094 0.155 0.000 0.992 0.217 0.000
Female -0.069 0.089 0.442 -0.181 0.125 0.147
Age 18-24 -1.062 0.153 0.000 -0.801 0.206 0.000
Age 65+ -0.399 0.127 0.002 -0.967 0.181 0.000
Resident in Northeastern US 0.563 0.121 0.000 0.501 0.170 0.003
Resident in Midwestern US -0.422 0.120 0.000 -0.584 0.167 0.000
Resident in Western US 0.252 0.112 0.024 0.074 0.157 0.634
Education graduate school or higher 0.240 0.096 0.013 0.344 0.135 0.011
Income (in $10k) 0.000 0.000 0.000 0.000 0.000 0.036
Number of adults -0.129 0.056 0.021 -0.238 0.078 0.002
Number of workers 0.152 0.052 0.003 0.165 0.073 0.023
Number of children 0.136 0.069 0.048 0.154 0.096 0.111
Number of vehicles 0.234 0.052 0.000 0.308 0.073 0.000
Full-time employed -0.614 0.107 0.000 -0.606 0.149 0.000
Married -0.274 0.108 0.011 0.001 0.152 0.996
Divorced -0.608 0.159 0.000 -0.334 0.220 0.129
No. of observations 1,004     1,004    
Dispersion Parameter (ρ): 0.693     0.335    
McFadden's R2: 0.198     0.114    
Likelihood ratio test (χ2) 231     110    
Prob > χ2 0.000     0.000    
2 x log-likelihood -4,635     -3,719    
Figure 1
Figure 1.Practical significance of variables in the trip frequency models

The business trip frequency model results show that more workers in the household and higher employment status (full-time employment versus part-time or unemployed) led to more LD business trips in 2019 and 2020. When unaffected by COVID-19, people age 25 to 64 living in the Western US with higher annual income were predicted to make more LD business trips. People with more children were predicted to make more LD business trips, to pursue higher household income, although taking care of children may often make one forsake business travel plans. The pandemic model (year 2020) shows fewer LD business trips made in general (as seen from the drop in the intercept), but females, people age 64 or younger, and those married people were less impacted. The most practically significant variable in the pre-pandemic model was residence in the Western US, but it became less practically significant during the pandemic as people worked from home. Most of the variables were less practically significant in the pandemic model; however, the variable of people age 65 or older was more significant, because they were more vulnerable to COVID-19, leading to fewer LD business trips.

For the non-business model, people aged 25 to 64 years made more non-business LD trips in 2020 compared to other age groups, as they have more needs in LD trip-making for non-business purposes, such as visiting family and friends and for recreation. A higher education level, more vehicles owned, and higher household income also led to more LD non-business trip-making. According to the practical significance in the non-business model (Figure 1b), the variable of people age 65 or older was less significant in LD non-business trip-making compared to business purposes, because these peopfle still need to engage in recreational activities or visiting family even if their travel was greatly impacted by the pandemic. An increase by one standard deviation ($51k) of household income led to a 26% increase in pre-pandemic LD non-business trip counts, but only led to a 16% increase during 2020, likely because of people’s unwillingness to make LD non-business trips when exposed to health risks. Interestingly, being full-time employed had a positive impact on LD trip frequency for business purposes, while it had the opposite impact on LD trip-making for non-business purposes.

Results from this survey have shown that COVID-19 has greatly impacted LD travel behavior, for both business and non-business purposes. The pandemic raises the need to gather global LD trip patterns, while policies regarding vaccine and travel restrictions across different states and nations are also fundamental to capture future LD travel patterns.

Acknowledgements

The authors thank the Texas Department of Transportation (TxDOT) for financially supporting this research, under research project 0-7081, “Understanding the Impact of Autonomous Vehicles on Long-Distance Travel Mode and Destination Choice in Texas”. The authors also thank Jade (Maizy) Jeong for her excellent editing and submission support.