Evaluating COVID-19’s impacts on Puerto Rican’s travel behaviors

Figures & data

Figure 1. Monthly ODT flows between Puerto Rico and the US.

Figure 2. Difference in travel flows from Puerto Rico to the contiguous US calculated as pre-pandemic flow minus peri-pandemic flow.

Table 1. Official definitions of the chosen variables from the CDC’s SVI (Flanagan et al. 2011) and how they impact travel decisions.

Variable	Definition and Impact on Migration
Below Poverty	Definition: Percent of population living below the federally defined poverty line (varies by size and age composition of households) Impact: Migration is a coping strategy for escaping poverty and insecurity (e.g. Siddiqui 2012; Snel and Staring 2001).
Uninsured	Definition: Percentage uninsured in the total civilian noninstitutionalized population estimate Impact: Improved insurance opportunities serve as pull factors in migration decisions (e.g. Simpson 2017)
Aged 65 or Older	Definition: Percent of the population 65+ years old Impact: Elderly typically need more in-house care and will migrate to be closer to adult children or seek better healthcare options (e.g. Ostrovsky 2004; Weeks, Keefe, and Macdonald 2012)
Civilian with a Disability	Definition: Percent of population 5+ years old living with a disability (excluding those in an institution) Impact: Similar to the elderly population, civilians with disabilities will migrate for better healthcare options (e.g. Burns 2020)
Single-Parent Households	Definition: Percent of households with one parent and children under 18 years old Impact: While smaller households tend to be more mobile, single-parent households statistically have lower socioeconomic status. The responsibility of dependents is put on one parent, lessening their migration ability (e.g. Flanagan et al. 2011).
Crowding	Definition: Percent of households with more people than rooms Impact: Overcrowding since the pandemic started has shown people moving from densely built areas to more spacious homes (e.g. Czarnecki, Dacko, and Dacko 2021).
No Vehicle	Definition: Percent of households with no vehicle Impact: A household with a vehicle is statistically more likely to migrate for a disaster than a household without (e.g. Gudishala and Wilmot 2012; Pedro 2006).

Table 2. O-C model-generated coefficients for the pre-pandemic and peri-pandemic periods.

Variable	Pre-Pandemic		Peri-Pandemic (cases)		Peri-Pandemic (deaths)
Destination Pop	1.422	***	1.283	***	1.327	***
Euclidean Distance	−3.151	***	−2.564	***	−2.784	***
% PR Pop at D	0.316	***	0.466	***	0.395	***
COVID-19 (new cases)		NA	1.228	***		NA
COVID-19 (new deaths)		NA		NA	0.212	***

Significance codes include ‘***’ for p-value < 0.001.

Table 3. D-C model-generated coefficients for the pre-pandemic and peri-pandemic periods.

Variable	Pre-Pandemic (First Leg)		Pre-Pandemic (Euclidean Distance)		Peri-Pandemic (First Leg)		Peri-Pandemic (Euclidean Distance)
Origin Pop	1.050	***	1.097	***	1.125	***	1.136	***
Euclidean Distance		NA	−1.760	***		NA	−1.766	***
First Leg Distance	−0.202	***		NA	−0.063	***		NA
Below Poverty	−0.460	***	−0.630	***	−0.484	***	−0.586	***
Aged 65 or Older	0.817	***	0.719	***	1.344	***	1.300	***
Civilian with a Disability	−0.008	*	−0.041	***	−0.058	***	−0.078	***
Single-Parent Households	−0.325	***	−0.121	***	0.069	***	0.158	***
Crowding	0.048	***	0.019	***	0.046	***	0.046	***
No Vehicle	0.045	***	0.187	***	−0.314	***	−0.247	***
Uninsured	−0.049	***	0.004		0.204	***	0.225	***

Significance codes include ‘***’ for p-value < 0.001, ‘*’ for p-value < 0.01, ‘’ for p-value > 1.

Figure 3. Coefficients of COVID-19 variables new cases (light blue bar) and new deaths (dark blue bar) by month. All coefficients were significant at the 0.001 level. Raw numbers of new cases and new deaths are included as light blue and dark blue lines, respectively. New case number was scaled down by a factor of 50 to show trends.

Figure 4. O-C model-produced coefficients for variables destination pop, % PR pop at D, and Euclidean distance, separated by month for pre-pandemic in gray and peri-pandemic in blue.

Figure 5. D-C model-produced coefficients for variables first leg and Euclidean distance separated by month for pre-pandemic in gray and peri-pandemic in blue.

Figure 6. D-C model-produced coefficients for variables origin pop and chosen SVI variables separated by month for pre-pandemic in gray and peri-pandemic in blue.

Flanagan, B. E., E. W. Gregory, E. J. Hallisey, J. L. Heitgerd, and B. Lewis. 2011. “A Social Vulnerability Index for Disaster Management.” Journal of Homeland Security and Emergency Management 8 (1): 1. doi:10.2202/1547-7355.1792.

 Web of Science ®Google Scholar

Siddiqui, T. 2012. “Impact of Migration on Poverty and Development.” Migrating out of Poverty. United Kingdom: University of Sussex, School of Global Studies. https://opendocs.ids.ac.uk/opendocs/bitstream/handle/20.500.12413/14824/wp2-impact-of-migration-on-poverty-and-development.pdf?sequence=1&isAllowed=y

 Google Scholar

Snel, E., and R. H. J. M. Staring. 2001. “Poverty, Migration, and Coping Strategies: An Introduction.” Focaal European Journal of Anthropology 38 (January): 7–22.

 Google Scholar

Simpson, N. 2017. “Demographic and Economic Determinants of Migration.” IZA World of Labor. doi:10.15185/izawol.373.

 Google Scholar

Ostrovsky, Y. 2004. “Life Cycle Theory and the Residential Mobility of Older Canadians.” Canadian Journal on Aging/La Revue Canadienne Du Vieillissement 23 (5): S23–37. doi:10.1353/cja.2005.0038.

 PubMedGoogle Scholar

Weeks, L. E., J. Keefe, and D. J. Macdonald. 2012. “Factors Predicting Relocation Among Older Adults.” Journal of Housing for the Elderly 26 (4): 355–371. doi:10.1080/02763893.2011.653099.

 Google Scholar

Burns, N. 2020. “Boundary Maintenance: Exploring the Intersections of Disability and Migration.” In Routledge Handbook of Disability Studies, edited by Nick Watson and Simo Vehmas. 2nd ed., 305–320. Oxon and New York: Routledge.

 Google Scholar

Czarnecki, A., A. Dacko, and M. Dacko. 2021. “Changes in Mobility Patterns and the Switching Roles of Second Homes as a Result of the First Wave of COVID-19.” Journal of Sustainable Tourism. 1–19. doi:10.1080/09669582.2021.2006201.

 Web of Science ®Google Scholar

Gudishala, R., and C. Wilmot. 2012. “Comparison of Time-Dependent Sequential Logit and Nested Logit for Modeling Hurricane Evacuation Demand.” Transportation Research Record 2312 (1): 134–140. doi:10.3141/2312-14.

 Google Scholar

Pedro, S. 2006. “Delineating Hurricane Vulnerable Populations in Orleans Parish, Louisiana.” LSU Master’s Theses, January. https://digitalcommons.lsu.edu/gradschool_theses/515.

 Google Scholar

Data availability statement

The data that support the findings of this study are openly available from the following locations: 1) US Census Bureau in [https://data.census.gov/cedsci/table?q=mobility&g=0400000US72&tid=ACSDT1Y2019.B07204PR&tp=true], reference number [B07204PR]; 2) CDC Case Task Force in [https://data.cdc.gov/Case-Surveillance/United-States-COVID-19-Cases-and-Deaths-by-State-o/9mfq-cb36], reference number [Centers for Disease Control and Prevention (009:20)]; 3) CDC/ATSDR Social Vulnerability Index in [https://www.atsdr.cdc.gov/placeandhealth/svi/data_documentation_download.html]; and 4) ODT Flow Explorer in [http://gis.cas.sc.edu/GeoAnalytics/od.html].