key: cord-0963055-nd690hcr authors: Yilmazkuday, Hakan title: COVID-19 Spread and Inter-County Travel: Daily Evidence from the U.S. date: 2020-10-22 journal: Transportation research interdisciplinary perspectives DOI: 10.1016/j.trip.2020.100244 sha: 44424c769762e32d3ceb7b489dea8e74d08f97ca doc_id: 963055 cord_uid: nd690hcr Daily data at the U.S. county level suggest that coronavirus disease 2019 (COVID-19) cases and deaths are lower in counties where a higher share of people have stayed in the same county (or travelled less to other counties). This observation is tested formally by using a difference-in-difference design controlling for county-fixed effects and time-fixed effects, where weekly changes in COVID-19 cases or deaths are regressed on weekly changes in the share of people who have stayed in the same county during the previous 14 days. A counterfactual analysis based on the formal estimation results suggests that staying in the same county has the potential of reducing total weekly COVID-19 cases and deaths in the U.S. as much as by 139,503 and by 23,445, respectively. federal government has left such policy decisions to local governments. § Based on this background, this paper investigates whether inter-county travel within the U.S. has any implications for COVID-19 cases or deaths. This is achieved by using U.S. daily data at the county level covering the period between January 21th, 2020 and September 2nd, 2020. Inter-county travel is measured by using data from smartphone devices. Descriptive statistics suggest that both COVID-19 cases and deaths are lower in counties where a higher share of people have stayed in the same county (or a fewer share of people have travelled across counties) during the previous 14 days. Since descriptive statistics cannot control for any county-specific characteristics or time-specific changes that are common across counties, a formal investigation is achieved by using a difference-in-difference design, where county-fixed effects and time-fixed effects are controlled for. The estimation results suggest that if a person lives in a county where the average person has travelled less compared to the previous week, it is better for this person to stay in her county to reduce the possibility of catching COVID-19 as her county has lower COVID-19 cases or deaths due to other people in that county travelling less. However, if a person lives in a county where the average person has travelled more compared to the previous week, it is better for this person to travel as well (potentially to counties with lower COVID-19 cases) to reduce the possibility of catching COVID-19 as her county has higher COVID-19 cases or deaths due to other people in that county travelling more. The estimation results are further used to answer the following hypothetical question based on a counterfactual analysis: What would happen to the number of COVID-19 cases and deaths in each county if all people would stay in the same county? The results suggest that staying in the same county has the potential of reducing total weekly COVID-19 cases and deaths in the U.S. as much as by 139,503 and by 23,445, respectively. At the county level, staying in the same county has the potential of reducing COVID-19 cases between 2 and 209 across counties, and it has the potential of reducing county-specific COVID-19 deaths up to 35. It is implied that staying in the same county (i.e., travelling less across counties) would help fighting against COVID-19. These Merler, yPiontti, Mu, Rossi, Sun et al. (2020) who have shown that the travel restrictions implemented in China have mitigated the spread of COVID-19. The rest of the paper is organized as follows. The next section introduces the data set and methodology used. Section 3 depicts and discusses empirical results, while Section 4 concludes. Daily U.S. data on the cumulative number of COVID-19 cases and deaths at the county level have been obtained from New York Times. ** Daily data for inter-county travel have been borrowed from Couture, Dingel, Green, Handbury, and Williams (2020) . † † The latter data set has been constructed by using PlaceIQ data that describe smartphone devices "pinging" in a given geographic unit on a given day. Based on this information, once a certain number of smartphone devices are determined to be in a particular U.S. county on a particular day, the data set provides information on the share of these devices that have pinged in another U.S. county at least once during the previous 14 days. ‡ ‡ The combined sample covers the daily period between January 21th, 2020 and September 2nd, 2020 for 2018 U.S. counties. Daily data for inter-county travel are used to obtain information on staying in the same county (or travelling less across counties) during the previous 14 days. Formally, given that there is a certain number of smartphone devices pinged in county c on time t , let's denote the share of these devices that have pinged in county i at least once during the previous 14 days with cit p . Based on this notation, we consider the following definition for staying in the same county (or travelling less across counties) during the previous 14 days. The summation of shares of devices that have not pinged (even once) in any other county during the previous 14 days. ** The web page is https://github.com/nytimes/covid-19-data/commits/master. † † The web page is https://github.com/COVIDExposureIndices. ‡ ‡ As detailed in Couture, Dingel, Green, Handbury, and Williams (2020) , although PlaceIQ data cover a significant fraction of the U.S. population, differences in smartphone ownership may result in unrepresentative samples; e.g., older adults are less likely to own smartphones, making smartphone-derived samples unbalanced across age groups. to have a counterfactual analysis below, where we will ask the following question: What would happen to the number of COVID-19 cases and deaths in each county if all devices would stay in the same county? For visual evidence, the treatment group is constructed as counties that have experienced a certain degree of an increase in , ct S , whereas the control group is constructed as the other counties. To consider seasonality by construction, we work with weekly changes. In particular, first, for each county, we first calculate weekly changes in , Specifically, . Therefore, we consider how much each county is close to those other counties experiencing a certain increase J o u r n a l P r e -p r o o f in their , ct S measures. When the number of COVID-19 cases in the U.S. are considered, the visual evidence based on travelling across counties ( , ct S measures) is provided in Figure 1 and summarized in Table 1 The corresponding historical patterns over time for the average COVID-19 cases (across counties) are given in Figure 1 . As is evident, independent of the threshold considered, less-travelling counties have experienced lower number of COVID-19 cases compared to more-travelling counties in the U.S., and the difference between these treatment and control groups gets higher for higher threshold values (as consistent with Table 1 ). The results of a similar visual investigation for the number of COVID-19 deaths based on travelling across counties ( , ct S measures) are given in Figure 2 and summarized in Table 2 . As is evident in Table 2 Table 2 ). The visual evidence provided so far does not control for any county-specific characteristics or time-specific changes that are common across counties. Moreover, the effects of staying in the same county (or travelling less across counties) may be asymmetric between counties depending on the sign of the potential of having COVID-19 in that county due to travelling more. In order to capture these additional details, we achieve a formal investigation based on the following difference-in-difference specification: Once Equation 2 is estimated, we further use the corresponding results to ask the following hypothetical question as briefly described above. Journal Pre-proof The results of estimating Equation 2 are given in would reduce inter-county travel, total number of COVID-19 cases and deaths can be reduced (as we analyze more during the counterfactual investigation, below). As is also evident in week, it is better for this person to travel as well (potentially to counties with lower COVID-19 cases) to reduce the possibility of catching COVID-19 as her county has higher COVID-19 cases or deaths due to other people in that county travelling more. What would happen to the number of COVID-19 cases and deaths in each county if all devices would stay in the same county? The answer to this hypothetical question is given in A counterfactual analysis based on the formal estimation results further suggests that staying in the same county has the potential of reducing total weekly COVID-19 cases and deaths in the U.S. as much as by 139,503 and by 23,445, respectively. At the county level, staying in the same county has the potential of reducing COVID-19 cases between 2 and 209 across counties, and it has the potential of reducing county-specific COVID-19 deaths up to 35. It is implied that staying in the same county (i.e., travelling less across counties) would help fighting against COVID-19. Although the investigation has been achieved at the county level, the results highly support several stay-at-home orders implemented by alternative layers of government in the U.S., especially during March and April 2020. Notes: ** and *** represent significance at the 1% and 0.1% levels. Standard errors are in parentheses. Human mobility networks, travel restrictions, and the global spread of proof pandemic The roles of transportation and transportation hubs in the propagation of influenza and coronaviruses: a systematic review A familial cluster of pneumonia associated with the 2019 novel coronavirus indicating person-to-person transmission: a study of a family cluster Human mobility and the spatial transmission of influenza in the United States Measuring Movement and Social Contact with Smartphone Data: A Real-Time Application to COVID-19 On the dynamics of SEIRS epidemic model with transport-related infection Human Mobility Restrictions and the Spread of the Novel Coronavirus (2019-nCoV) in China The effect of human mobility and control measures on the COVID-19 epidemic in China The association between international and domestic air traffic and the coronavirus (COVID-19) outbreak Outbreak dynamics of Table 3 .