MMobility-based contact exposure explains thedisparity of spread of COVID-19 in urbanneighborhoods
Rajat Verma , Takahiro Yabe , and Satish V. Ukkusuri Lyles School of Civil Engineering, Purdue University, West Lafayette, 47906, U.S.A. * Corresponding author: [email protected]
The rapid early spread of COVID-19 in the U.S. was experienced very differently by different socioeconomic groups andbusiness industries. In this study, we study aggregate mobility patterns of New York City and Chicago to identify therelationship between the amount of interpersonal contact between people in urban neighborhoods and the disparityin the growth of positive cases among these groups. We introduce an aggregate
Contact Exposure Index (CEI) tomeasure exposure due to this interpersonal contact and combine it with social distancing metrics to show its effecton positive case growth. With the help of structural equations modeling, we ﬁnd that the effect of exposure on casegrowth was consistently positive and that it remained consistently higher in lower-income neighborhoods, suggesting acausal path of income on case growth via contact exposure. Using the CEI, schools and restaurants are identiﬁedas high-exposure industries, and the estimation suggests that implementing speciﬁc mobility restrictions on thesepoint-of-interest categories are most effective. This analysis can be useful in providing insights for government ofﬁcialstargeting speciﬁc population groups and businesses to reduce infection spread as reopening efforts continue to expandacross the nation.
Keywords — COVID-19, coronavirus, human mobility, contact, exposure, social distancing, stay-at-home,disparity, socioeconomic factors
The rapid worldwide spread of the novel coronavirus disease (COVID-19) has caused signiﬁcant distress to citizens andgovernments worldwide and has warranted unprecedented control measures on human mobility such as travel bans,mandatory quarantines, and stay-at-home/shelter-in-place orders. In the United States, like many other countries, theserestrictive policies have incurred huge economic losses , leading to experts showing interest in targeted policy-makingbased on differential effects of the virus on different subject and activity types. These policies may be either supportive,such as the CARES Act relief based on household income , or restrictive, such as heightened restrictions on somebusiness activities more than others, such as schools, bars, concerts, sports events, and indoor dining. Similarly, in somecases, speciﬁc neighborhoods within cities may be classiﬁed as “coronavirus hotspots” that are subjected to additionalprecautions and heightened restrictions, such as the New Rochelle community in the NYC metropolitan region in March2020 .It has been shown that these policies have had differential impacts on the livelihood and health of differentsocioeconomic groups within cities and states, with certain factors such as income , education level , and race standing out as the principal discriminants, in addition to age which is inherently strongly correlated with highersusceptibility to infection by and morbidity due to the coronavirus. In considering targeted policies, such as targetingurban neighborhoods, it then becomes crucial to understand how these different social groups may respond to suchpolicies in terms of mobility and the growth of the disease.A crucial hindrance to this approach is a lack of high-ﬁdelity publicly available epidemiological data related tothe disease. State governments and popular COVID-19 trackers such as the Covid Tracking Project and the JHUCoronavirus Research Center generally provide the data at the county level. Epidemiological data are scarce forsmaller spatial units such as zip code tabulation areas (ZCTAs) and have been only recently made publicly accessibleand by only a few regional governments such as those of NYC , Illinois , and Ohio .Many studies that have exploited county-level data have shown evidence of the effectiveness of these social a r X i v : . [ ec on . GN ] F e b istancing measures and related mobility reduction in reducing the rate of daily new COVID-19 positive cases, bothin the U.S. as well as in other countries . However, there is one major concern associated with the currentresearch focusing on the relationship between aggregate mobility and the spread of COVID-19 that we address in thisstudy. In the absence of human movement data at the individual scale, most studies consider mobility as an aggregateentity measured by the total number of trips between counties or other large regions or distance-based measures .Although overall movement trafﬁc ﬂow is a reasonable proxy of social distancing among travelers as a whole betweenregions, it suffers from two crucial limitations. First, social distancing inherently involves physical interaction betweenindividuals which is not directly captured by trafﬁc ﬂow and distance traveled. Travelers who travel long distances aloneor in small groups with little exposure to contact with others, such as in private vehicles, are misclassiﬁed as equallypotential vectors of the virus as those who travel in close proximity with large groups, such as public transit during thepeak hours . Second, such measures may not be a good indicator of the contact exposure travelers are subjected to attheir destinations, where they are likely to spend more time in proximity of other visitors than the means of travel. Thisincludes the number and physical spacing of other visitors at the destination as well as their common dwell time incontact with each other. These are important factors to be taken into account since the Centers for Disease Controland Prevention speciﬁcally highlights via its famous 6-feet-15-minutes rule for a close contact . While recent studieshave considered the distribution of trips by dwell time and crowd density based on non-residential square footage atcertain trip destinations, these have been studied in isolation as time series trends but not as a comprehensive measureof exposure.In this work, we address these two limitations - lack of a contact exposure-based mobility measure, and mobilityanalysis at the ZCTA level - in understanding the differential impact of mobility restrictions on socioeconomic groupswithin cities and states. We leverage anonymized origin-destination foot trafﬁc movement patterns based on mobilephone GPS records and aggregated by SafeGraph Inc. at the level of urban neighborhoods (as ZCTAs) and the businesscategory/industry of the trip destinations, hereby referred to as places of interest (POIs). Although there are signiﬁcantconcerns with using mobile phone GPS data, such as low penetration rate and privacy and data protection issues ,they have been nonetheless used both in general mobility analysis as well as speciﬁcally for COVID-19 .Based on the insights obtained from the studies exploiting county-level trip/distance measures, we hypothesizethat the number of positive coronavirus cases in residential neighborhoods is positively correlated with the amountof contact exposure their residents are subjected to when they travel outside their neighborhoods. Furthermore, wehypothesize that this difference in aggregate contact exposure emanates from, among other things, the difference inseveral socioeconomic factors of the residents. We test this causal pathway using structured equations models basedon a panel of daily mobility and case growth data of ZCTAs in New York City (NYC) and the Chicago metropolitanarea up to June 2020. Finally, we analyze the contribution of exposure to 12 major trip destination types (e.g., schools,hospitals, and restaurants) to the overall exposure to understand which industries could be targeted for tighter mobilityrestrictions in the event of a growth spurt in COVID-19 cases. In doing so, we also study the variation in the negativeeffect of income on exposure felt in these industries. Results
Lower income groups had higher exposure and more cases
In this section, we analyze the heterogeneity of infection spread by socioeconomic status in NYC and Chicago which hadbeen some of the worst-hit cities by COVID-19 during the ﬁrst wave of the pandemic . We quantify this heterogeneityalong the dimension of mean household income of the ZCTAs. We also use an unweighted gravity model-like approachto distribute the exposure (CEI) from each POI to the ZCTA of its visitors on each day and then aggregate over all POIsto get the exposure level of each ZCTA. Fig. 1 shows the cumulative number of positive cases and the cumulative CEIof the two study cities in the week of 20-26 April, since the earliest date for which reliable ZCTA-level epidemiologicaldata are available for Chicago is April 18. The dots at the centroids of the ZCTAs depict their income class, given bythe quintile of the average household income distribution of the ZCTAs as of 2017.At ﬁrst glance, it can be observed that lower income regions (such as The Bronx in NYC and southern Chicago) haddisproportionately more positive cases than their higher income counterparts in this week. They also experienced higherexposure to contact as measured by total CEI during that period. While cases were relatively more evenly distributed inNYC, with some peaks in neighborhoods in Queens and King Counties, they were much more concentrated in Southand downtown Chicago, albeit much fewer than NYC in general. Moreover, the virus transmission started to decline inChicago about two weeks after this week, while it had already started declining since at least early April (see Fig. 2A).On the contrary, though, mobility began increasing in April in both the cities after the initial phase of mobility plummet Map of study cities showing total cases and CEI. ZCTAs of the two study cities showing ( A ) total cases and ( B )total contact exposure index (min/ft) during 20-26 April. For reference, the quintile class of the mean household incomeof the ZCTAs are also shown as colored centroid dots. following the issuance of stay-at-home orders on March 20 in NYC and Chicago , with total exposure to contactinitially decreasing rapidly (Fig. 2B) and more people spending more time at home (Fig. 2C and D). Figure 2.
Trends of cases and mobility measures. Daily variation of ( A ) new positive cases, ( B ) contact exposure index,( C ) fraction of devices registered as staying at home all day, ( D ) median time spent at home. The light shaded curvesdenote the daily trends while the dark ones depict their 7-day forward-shifted moving average. The vertical lines in panel A represent the ﬁrst dates of available data of the number of cases. On closer inspection, the relationship between exposure-based mobility (measured by CEI), the spread of the virus,and income becomes more evident when looking at the total exposure subjected to the population of neighborhoods(see Fig. 3A). However, even after controlling for the population of the neighborhoods, we ﬁnd that exposure per capitais strongly associated with cases per capita (Fig. 3B), although the strength of this correlation reduces after controllingfor population. For reference, the distribution of population of the ZCTAs across these income groups is shown in Fig.3C. This effect likely occurs due to a higher proportion of visitors belonging to lower-income neighborhoods after theimposition of the stay-at-home order in NYC (Fig. 3D). This observation supports the idea that lower income peoplewere more susceptible to infection after the stay-at-home orders came into effect in these cities primarily because of thenature of their professions, mainly a higher representation of jobs requiring on-site work and/or belonging to essentialservices such as nursing, grocery store operations, etc. Econometric Modeling
We hypothesize that the difference in the caseload of lower-income neighborhoods can be explained by the difference inthe amount of exposure to contact they were subjected to during the early phase of the lockdown. We highlight theimportance of measuring this exposure with CEI and other social distancing metrics instead of relying on the variationof number of trips since the latter may misrepresent the exposure by unnecessarily counting solo trips and discountingthe interaction with others at the destination.Here, we test this hypothesis by testing the strength of the causal path of household income to virus transmission
Relationship between cases, exposure, and income. ( A ) New positive cases versus total exposure (CEI) in theZCTAs of NYC in the week of 6-12 April, differentiated by the ZTCA income quintiles, along with the probability densitydistribution of ZCTAs by income class; ( B ) same as A but with values divided by ZCTA population. For reference,distributions of ( C ) population of ZCTAs and ( D ) the number of visitors in this week across these income classes are alsoshown, along with the intervals covered by these classes. through a latent measure of exposure to social contact using a structured equations model (SEM). We specify thisexposure measure to be latent so that it can take into account the effects of social interaction at places of commercial aswell as non-commercial activity and compensate for the shortcomings of the used mobility variables. While exposureat commercial places is captured reasonably well with CEI at POIs (which include most major places of commercialactivity except ofﬁces of private ﬁrms), the two social distancing measures - Prop Home and
Time Home , estimate theexposure due to travel outside home without considering social interaction.We develop daily SEMs where for each day t , a causal pathway is assumed from 6 static socioeconomic variables,particularly mean household income, to the daily number of new cases via latent exposure measured by the dailymobility variables (mutually correlated). Models of similar design have been used to show the impact of inter-countytravel ﬂow on case growth rate , though without considering contact exposure as a facilitator. For more details of themodel structure used, see the Materials and Methods section. The variables are described in Table 1.The parameter estimates of the two main relationships of interest in the daily SEMs of the two cities are shownin Fig. 4, along with the standard error of these estimates. The coefﬁcients β S [ ] measuring the effect of income onexposure (panel A) on exposure are consistently negative for both the cities. This implies that residents of lower incomeneighborhoods in these cities remained more likely to coming in close contact with other individuals throughout thestudy period. This effect is higher in the case of Chicago, although the large ﬂuctuations in this effect are not correlatedwith any remarkable shift in other variables, such as lifting of the lockdown or a sudden and brief change in publicresponse to COVID-19 that could have triggered this change. This difference in effects between NYC and Chicagocould also emanate from the already large pool of infected people and higher mobilization of resources in NYC due toits uniquely intense peak of cases in mid-March .A similar consistency is also observed in the effect of exposure on daily new cases where an increase in exposure islinked with a corresponding increase in the number of cases (Fig. 4B). In this case, however, the effect is higher inthe case of NYC compared to Chicago, implying that NYC was more sensitive to mobility changes in terms of virustransmission than Chicago during the initial months after lockdown. After controlling for unobserved variables in thesemodels, one could interpret this in this way - even if household income in NYC is not as considerable an indicator ofexposure to contact as in Chicago, contact exposure contributed more substantially to the growth of cases in NYC thanin Chicago.These observations provide the core insight for understanding the causal mechanism of trip destination-based Description of the independent, target, and latent variables used in the dailystructured equations models, along with their symbols and ranges.
Category Name Description Range
Socioeconomic, S i Income
Natural logarithm of mean household income of ZCTA i [ , ∞ ) Low Edu
Fraction of population having a high school diploma or less [ , ] Poor
Fraction of population classiﬁed as living below the povertyline [ , ] Age Fraction of population aged 65 years and above [ , ] Black
Fraction of population which identiﬁes as African American(monoracial) [ , ] Transit
Fraction of population whose primary mode of commute towork is public transit (buses, monorail, or subway) [ , ] Mobility, M i , w CEI E i , w Weighted total contact exposure index (min/ft) of POI visitorsliving in ZCTA i in the time window w = [ t − , t ) , transformedwith the mapper f ( x ) = ln ( + x ) [ , ∞ ) Time Home T i , w Fraction of day spent by the mobile devices within theGeohash7 of their owners’s home in ZCTA i on day t ,averaged over the window w [0,1] Prop Home P i , w Proportion of mobile devices registered as staying within thehome Geohash7 of residents of ZCTA i all day on at leastone day in the window w [0,1] Devices N i , w (not used) Number of mobile devices belonging to residents of ZCTA i on day t ; not used directly in the model but only to aggregate Prop Home and
Time Home over the window w [ , ∞ ) Health
Cases y i , t Number of new positive cases in ZCTA i on day t , alsotransformed with f ( x ) [ , ∞ ) contact exposure on the course of COVID-19 in the early phase of lock-downs in these cities. In the next section, wediscuss how this exposure varied across different destination types, which could be used to identify the industries activein helping spread COVID-19 faster. Exposure by Destination Types
To better understand the characteristics of trips that contribute more to exposure to contact, we next discuss the traveltrends in NYC and Chicago to POIs of different industries. The publicly available Google Community MobilityReport has been commonly used to study the differences in travel behavior across trip categories . However, itonly provides data at the state or county level and for a select travel categories, such as home, work, groceries, etc.,meaning there is limited opportunity to explore speciﬁc industries of interest, such as bars and hospitals. The SafeGraphmobility patterns are provided at the POI level, so they can be used to study these categories in detail, such as in .We focus on 12 popular industries (by daily visits) in the study cities and label them according to their industrycodes as per the North American Industry Classiﬁcation System (NAICS). The trends of the total daily exposure (CEI)across these industries are shown in Fig. 5, along with their NAICS codes and total number of visits to their POIs in theentire city (excluding lone hourly visits). In this ﬁgure, it can be seen that all of the industries experienced a drasticdecline after the declaration of emergency in the two cities, with many categories falling close to zero exposure, such asschools and malls immediately after the lockdown (stay-at-home rule). Since then, exposure has increased in hospitalsand at fast food places but has largely remained negligible compared to before emergency. While schools and ﬁtnesscenters have seen the biggest plummet in exposure, supermarkets have seen the lowest decline after a brief weekendsurge in Chicago, likely due to panic buying.There are a few interesting shifts in travel behavior across the two cities. CEI in the Chicago metropolitan area hadbeen lower on average than NYC at supermarkets, fast food places, and gas stations prior to the shutdown, but thispattern started reversing afterwards. This could simply be a consequence of the severity of enforcement of lockdownpractices in NYC, with several reports discussing the severe punitive actions being taken against social distancingviolators during this period. Estimates of important parameters of daily SEMs. Daily series of coefﬁcient estimates of the two SEMrelationships of interest for Chicago and NYC shown as 7-day moving averages: ( A ) effect of income on exposure, and( B ) effect of exposure on new cases. The shaded region indicates the region spanned by estimate ± standard error. Contribution of Industries on Exposure
Though Fig. 5 provides an overview of mobility patterns across the major industries considered here, an importantfactor in the consideration of categorical restrictions is the contribution of these industries to the overall exposure tocontact at a macroscopic level. This is clariﬁed in Fig. 6 where panels A and B show the proportion of CEI comingfrom POIs in the 12 important industries. The differences before and after the emergency declaration are evident inmany categories. An interesting shift in the pattern of contact exposure is the near eradication of weekly recurringpatterns in both the cities, primarily achieved through the closure of services that typically show strong trafﬁc variationbetween weekdays and weekends, such as schools, daycare centers, ﬁtness centers, and eating places.Schools offer a particularly interesting case in point. Public schools in NYC were closed on March 16, but manyprivate schools and school districts had already started closing a week ago. Prior to closure, schools exhibited someof the highest exposure, both in terms of total CEI (Fig. 6A) and average CEI per visit (Fig. 6C). This makes sensegiven that visits to schools typically have much higher dwell time (typically 4-5 hours) than other POIs and also havemore visitors in general (see Fig. 5). The drop in exposure following school closure is even starker in Chicago whererecurring periods of high CEI vanished almost overnight close to its date of issuance of the stay-at-home order. Theseobservations support the decision of the public authorities of closing schools on the grounds of exposure to contact.The full-service restaurant industry also stands out as being the dominant destination type for visitors over the entirestudy period for both the cities. Although this industry has not seen a decline in exposure as sharp as schools anddaycare centers, it has had the most impact in the total reduction of exposure. This was likely facilitated by the differingdegrees of closure following lockdown, with most restaurants remaining fully shut while a few others provided outdoordining services .The results for these two industries provide afﬁrmation for the actions taken by public authorities in exercisingspecial restrictions on them. However, these decisions have had implications on the disparity of contact exposure acrossneighborhoods of different income levels. This difference by industry is highlighted in Fig. 7. This ﬁgure shows therelationship between POI industry and contact exposure considering the income level of the people who visit thesePOIs and reafﬁrms the sharp decline in the exposure (CEI) in schools, ﬁtness centers, and bars, especially in Chicago.We chose 4 weeks representative of the different phases of mobility restrictions in the study period to summarizethe evolution of this disparity - mid-February (pre-lockdown), mid-March (just after lockdown), late April (a monthafterwards), and early June (beginning of reopening).A clear pattern in this ﬁgure is a consistent ordering of CEI with respect to income classes across all the categoriesand both the cities. Even though CEI declined very sharply after the implementation of stay-at-home, this order did notchange much. Interestingly, the ratio of the exposure generated by the lower-income neighborhoods to that generatedby the higher-income ones did not change substantially in Chicago but increased substantially in NYC. This can beinferred by noting the difference in the width of the bands across the industries at the left end (pre-lockdown) with theright end (lockdown and reopening), which on a linear scale represents the ratio of CEI. It could be argued that a stricterlockdown in NYC could have triggered a more polarized response from the public partly attributable to the inability ofthe lower-income people to stay at and work from home.These observations provide new and conﬁrm already accepted insights pertaining to mobility and spread ofCOVID-19, such as the rise of socioeconomic disparity in cities during at least the early period of the pandemic and the Trends of CEI by industry. Daily trends of CEI to POIs of 12 popular industries in NYC and Chicago, shown as7-day moving averages. The lighter shaded curves are the daily trends. For reference, the 6-digit NAICS code and thetotal number of multi-person visits to POIs between February 1 and June 28. The dates of two main mobility restrictivepolicies in these cities are also shown. role of special restrictions on certain types of places.
Social distancing policies like stay-at-home orders and closure of many services have caused widespread decline inoverall mobility since mid-March due to the spread of COVID-19 in the U.S. While research has shown that theserestrictions have been associated with an increase in socioeconomic disparity among urban neighborhoods, little workhas been done on understanding the mechanism of this change. Also, while current research in this respect has oftenrelied on macroscopic mobility measures like population ﬂow and distance travelled, there has been limited work whichseeks to understand the effect of mobility by exploiting the knowledge that one of the most important causes of thespread of this disease is coming in close contacts with an infected person.In this study, we attempt to create the relationship bridge between human mobility involving high exposure toCOVID-19 and the effect of the rapid change in this mobility on the rise of socioeconomic disparity in U.S. cities byanalyzing the contact exposure-based mobility patterns of Chicago and New York City in the ﬁrst four months of thewidespread outbreak of the pandemic. Based on aggregate mobile phone-based mobility data provided by SafeGraphInc., we develop a Contact Exposure Index (CEI). This is an aggregate mobility metric based on three important factorsassociated with the idea of socio-physical contact - the total number of people who visit a place within the city, the areaover which the visitors are spread over, and the duration of their stay there, with a special consideration of the scheduleof their visits.We observe that income is a consistently strong indicator of contact exposure measured by CEI in both the cities.Recognizing that CEI does not capture socio-physical interaction at places not classiﬁed as places of interest (POIs),we conceptualize an abstract notion of contact exposure measured by CEI in combination with two social distancingmetrics that we believe affect contact exposure - proportion of the mobile phone-tracked population staying at home allday, and the amount of time they spend at home. We establish the negative effect of mean household income of zipcode areas on this latent contact exposure and in turn the positive effect of exposure on the number of COVID-19 casesusing a time series of structured equations models.
Contribution of CEI by industry. Proportion of total daily CEI attributable to the 12 industries of interest in thestudy cities ( A ) before and ( B ) after the declaration of emergency. The other industries contribute the remainder of theCEI, denoted by the empty region in the chart areas. ( C ) Monthly change in average CEI per visit in the two cities, withMarch divided into two parts (before and after March 15). We then attempt to explain the composition of contact exposure by the destination categories (industries) of thetrips generating that exposure, measured by CEI. We observe that heightened restrictions on mobility to POIs ofcertain categories, such as schools and restaurants, have contributed substantially to the decline in the overall exposureto socio-physical contact in these cities, while the effect of closure of bars has limited contribution to this decline.This lends support to the idea of industry-speciﬁc targeted lockdown policies that the government ofﬁcials have beenimplementing throughout the pandemic. Finally, we also observe that the disparity in contact exposure by income classconsiderably increased over all of the important industries after lockdown in Chicago but not much in New York.Given the practical importance of these insights, we also recognize the numerous limitations with this macroscopicapproach of quantifying exposure to COVID-19 manifested from close interpersonal contact. First, the contact exposureindex we propose is based on assumptions about the spatiotemporal positioning of visitors within POIs that are highlyideal and likely uncommon. Second, the scale of this measure is highly dependent on the true number of visitors atPOIs which under the currently available data is a valid concern due to issues related to low representative coverage ofthe mobile devices tracked by SafeGraph.Having said that, we assert that this measure nevertheless provides more pertinent information about COVID-19transmission and is more comprehensive than ﬂow and distance-based measures and should be pursued as a tool ofmonitoring the progress of policies pertaining to mobility restriction, especially now that cities have begun reopeningdespite the pandemic soaring in the U.S. We hope to extend this study to provide a sound basis to the validity andpractical applications of this measure and the insights in this study in the future.
Materials and Methods
Two mobility datasets were obtained from SafeGraph Inc. whose foot trafﬁc records have been used in many studiesrelated to mobility during the COVID-19 pandemic . The ﬁrst dataset provides information about trips originatingfrom different CBGs as deﬁned in the American Community Survey (ACS) of 2013-2017. It includes relevantinformation such as the number and duration of devices staying in their home CBGs and their destination CBGs.The second dataset contains information of about 4 million POIs across the U.S., including their unique 6-digitNAICS code (deﬁning the industry of the POIs), ﬂoor area, enclosing CBG, hourly count of trips to these POIs, weeklydistributions of trip distance, weekly trip dwell time, and origin CBGs of the visitors. We excluded the POIs located
Phase comparison of CEI by industry and income. Variation of CEI generated by residents of neighborhoodsof 5 income classes by visiting POIs of the industries of interest in 4 speciﬁc weeks representing different phases of thestudy period in ( A ) NYC and ( B ) Chicago. inside hospitals (mostly fast food restaurants) because of classiﬁcation error due to the surge in visits to hospitals in thestudy cities during the peak of the pandemic. Epidemiological Data
Information about daily new tests, positive cases, deaths, and hospitalizations due to COVID-19 was obtained from therespective government health department websites - NYC and Chicago . We considered the Chicago metro region asthe ZCTAs included in 5 main counties - Cook, DuPage, Lake, Kane, and Will, totaling 253 ZCTAs. The data for thesecounties were derived from the Illinois dataset which has data available starting from 18 April 2020. The NYC healthdataset spans 178 ZCTAs across the ﬁve boroughs in the NYC area, and has daily updated data starting from 3 April2020. The resultant dataset has missing information about tests between 18 May and 6 June, so testing rate data was notconsidered in this study. Socioeconomic Factors
The 2017 ACS was used to obtain socioeconomic variables of interest at the CBG level which were then aggregatedat the ZCTA level. A principal component analysis of these variables resulted in the selection of 6 main measures ofsocioeconomic standing which are listed in Table 1.
Exposure-based mobility was measured with three metrics - two social distancing metrics pertaining to residents’movement outside of their neighborhoods, and a POI-based contact exposure index (CEI).
Duration and Proportion of Stay at Home
We used two measures of the degree of compliance to the stay-at-home orders issued in NYC and Chicago inMarch 2020 using the dwell time composition of mobile devices -
TimeHome , which are describedin Table 1, considering them reasonable indirect measures of social distancing practices . These metrics aremeasured by SafeGraph based on their estimated assignment of device owners to their home neighborhoods andprovided at the CBG level, which we aggregated to the ZCTA level. For more details about these metrics, seehttps://docs.safegraph.com/docs/social-distancing-metrics. Contact Exposure Index (CEI)
We deﬁne a Contact Exposure Index (CEI) that estimates the amount of exposure subjected to an individual during theirvisit to a place of interest (e.g. school, hospital) and possibly come in close proximity with other visitors. It takes intoaccount three key attributes of such trip-making - the number of people coming in contact with each other, the spacingbetween them at the time of contact, and the duration of this contact. he ﬁrst component is directly measured by the number of trips to a given place in a given day and has been usedextensively in estimating the effect of mobility on community transmission . The other two components requiremicroscopic details for exact computation which are generally not available on a large scale. Hence, we approximatethese by making some general assumptions on the trips to POIs. These assumptions are that (i) visitors are uniformlyspread out on the ﬂoor area of the POI at any given time, so that they can be assumed to be arranged in a square grid,(ii) all visitors arrive at the POI at the beginning of the hour of their trip, (iii) as in the worst case, every visitor comesinto contact with every other visitor for the entire duration of their stay at the POI, which may vary individually.We deﬁne contact exposure index of a given POI in a given hour as the worst-case total contact duration of itsvisitors divided by the square root of the POI ﬂoor square footage. It is measured in minutes/foot. It can be easilyaggregated over higher scopes (such as POI-daily level or ZCTA-daily level) by simply summing over hours. Itsexpressions for the 3 scopes considered here are given below.POI-hourly: E p , h = τ p , h (cid:112) A p ; POI-daily: E p , t = ∑ h = E p , h ; ZCTA-daily: E i , t = ∑ p ∈ z E p , t (1)Here, τ p , h is the total contact duration (in minutes) of POI p (lying in ZCTA z ) during hour h of day t and A p is theﬂoor area of the POI p in squared feet, which excludes parking lots but may include unusable space such as for ﬁxtures. τ p , h is the sum of contact duration of each pair of visitors in hour h , given by the minimum of their dwell times, sinceboth visitors have to be physically present at the POI to come into contact with each other.This can be illustrated with an example. Suppose we have 6 persons (say, A-F) visiting a POI between 1:00 and2:00 PM with the following dwell times (minutes): A:10, B:10, C:20, D:20, E:20, F:40. According to our assumptions,visitors A and B come into contact with everyone else for 10 minutes, so the total contact duration of both A and B is10 ∗ =
50 min. Visitors C, D, E come into contact for 10 minutes with A and B and 20 minutes with everyone else,so their contact duration is 10 ∗ + ∗ =
80 min. Finally, F contacts A and B for 10 minutes and with C, D, and Efor 20 minutes, so its contact duration is 10 ∗ + ∗ =
80. Since each pair has to be counted once, the total contactduration is half of the sum of these individuals’ contact duration, i.e., 0 . ∗ ( + + + + + ) =
210 min.When the dwell time distribution is discrete, such as in the SafeGraph data, it has to be assumed that each tripin a given bucket has a dwell time equal to the representative point of that bucket. For a k -bucketed distribution, theexpression of total contact duration is τ p , h = k ∑ i = µ i n i (cid:32) n i − + k ∑ j = i + n j (cid:33) (2)Here, n i is the number of trips to POI p in hour h whose dwell time lies in the i th bucket, with µ i denoting therepresentative point of that bucket. It can be seen that visits in higher duration buckets dominate this measure, makingit more realistic in terms of the compounding effect of trip duration on contact exposure. It also follows from thisexpression that a higher value of k (corresponding to ﬁner intervals) provides a more accurate estimate of total contactduration. The dwell time distribution in the SafeGraph data is given by the k = [ , ) , [ , ) , [ , ) , [ , ∞ ) ,so we chose µ = [ . , . , , ] minutes. Structured Equations Model
The model form used in the daily SEMs is represented by the following system of equations. The variables are describedin Table 1. y i , t = α y + β · y i , t − + β η · η i , w + ε y i , t + µ y i + ν y t (3a) η i , w = α S + β TS S i + ε S i , t + µ S i + ν S t (3b) E i , w = t − ∑ z = t − E i , z = α E + · η i , w + ε E i , t + µ E i + ν E t (3c) P i , w = t − ∑ z = t − N i , z P i , z = α P + β P · η i , w + ε P i , t + µ P i + ν P t (3d) T i , w = t − ∑ z = t − N i , z T i , z = α T + β T · η i , w + ε T i , t + µ T i + ν T t (3e) or each day t , this form assumes a causal impact of static socioeconomic variables ( S i , all mutually correlated)on the total latent exposure, η i , w , measured by the daily mobility variables ( M i , w , all mutually correlated) which itselfinﬂuences the number of cases on that day, y i , t . The effects of unmeasured contributory factors, such as testing rateand human behavior (better hygiene, use of protective face masks, personal motivation to travel, etc.) are captured bythe number of cases in the neighborhood on the previous day, y i , t − . Also, we consider the total mobility of the past7 days ( w = [ t − , t ) ) as contributing to the growth of cases on day t instead of the exposure on that day, based on amanifestation period of 7 days for COVID-19 (5 days for incubation + 2 days for reporting). In these equations, ε , µ ,and ν respectively denote the random spatiotemporal, ﬁxed spatial, and ﬁxed temporal error terms. Data Availability
The COVID-19 cases data are available on the state department of health websites of NYC and Chicago References Thunström, L., Newbold, S. C., Finnoff, D., Ashworth, M. & Shogren, J. F. The Beneﬁts and Costs of Using SocialDistancing to Flatten the Curve for COVID-19.
J. Beneﬁt-Cost Analysis , 179–195, DOI: 10.1017/bca.2020.12(2020). US-Congress. The Coronavirus Aid, Relief, and Economic Security (CARES) Act (2020). Weill, J. A., Stigler, M., Deschenes, O. & Springborn, M. R. Social distancing responses to COVID-19 emergencydeclarations strongly differentiated by income.
Proc. Natl. Acad. Sci. United States Am. , 19658–19660, DOI:10.1073/PNAS.2009412117 (2020). Brough, R., Freedman, M. & Phillips, D. Understanding Socioeconomic Disparities in Travel Behavior during theCOVID-19 Pandemic.
SSRN Electron. J.
DOI: 10.2139/ssrn.3624920 (2020). Jia, J. S. et al.
Population ﬂow drives spatio-temporal distribution of COVID-19 in China.
Nature , 389–394,DOI: 10.1038/s41586-020-2284-y (2020). Webb Hooper, M., Nápoles, A. M. & Pérez-Stable, E. J. COVID-19 and Racial/Ethnic Disparities.
JAMA ,2466–2467, DOI: 10.1001/jama.2020.8598 (2020). The COVID Tracking Project (2020). https://covidtracking.com. Dong, E., Du, H. & Gardner, L. An interactive web-based dashboard to track COVID-19 in real time.
The LancetInfect. Dis. , 533–534, DOI: 10.1016/S1473-3099(20)30120-1 (2020). New York City Government. NYC Coronavirus Disease 2019 (COVID-19) Data (2020). github.com/nychealth/coronavirus-data.
Illinois Department of Public Health, Illinois COVID-19 Statistics. (2020). dph.illinois.gov/covid19/covid19-statistics.
Ohio Department of Health. Ohio COVID-19 Dashboard (2020). coronavirus.ohio.gov/wps/portal/gov/covid-19/dashboards/key-metrics/cases-by-zipcode.
Courtemanche, C., Garuccio, J., Le, A., Pinkston, J. & Yelowitz, A. Strong Social Distancing Measures In TheUnited States Reduced The COVID-19 Growth Rate.
Abouk, R. & Heydari, B. The Immediate Effect of COVID-19 Policies on Social Distancing Behavior in the UnitedStates.
SSRN Electron. J.
DOI: 10.2139/ssrn.3571421 (2020).
Andersen, M. Early Evidence on Social Distancing in Response to COVID-19 in the United States.
DOI: 10.2139/ssrn.3569368 (2020). Gatto, M. et al.
Spread and dynamics of the COVID-19 epidemic in Italy: Effects of emergency containmentmeasures.
Proc. Natl. Acad. Sci. United States Am. , 10484–10491, DOI: 10.1073/pnas.2004978117 (2020).
Lai, S. et al.
Effect of non-pharmaceutical interventions to contain COVID-19 in China.
DOI: 10.1038/s41586-020-2293-x (2020).
Badr, H. et al.
Social Distancing is Effective at Mitigating COVID-19 Transmission in the United States. medRxiv , 2020.05.07.20092353, DOI: 10.1101/2020.05.07.20092353 (2020).
Kraemer, M. U. et al.
The effect of human mobility and control measures on the COVID-19 epidemic in China.
Science , 493–497, DOI: 10.1126/science.abb4218 (2020).
Linka, K., Peirlinck, M., Sahli Costabal, F. & Kuhl, E. Outbreak dynamics of COVID-19 in Europe and the effect oftravel restrictions.
Comput. Methods Biomech. Biomed. Eng. , 710–717, DOI: 10.1080/10255842.2020.1759560(2020). Yabe, T. et al.
Non-compulsory measures sufﬁciently reduced human mobility in Tokyo during the COVID-19epidemic.
Sci. Reports , 1–9, DOI: 10.1038/s41598-020-75033-5 (2020). 2005.09423. Labonté-Lemoyne, É., Chen, S. L., Coursaris, C. K., Sénécal, S. & Léger, P. M. The Unintended Consequences ofCOVID-19 Mitigation Measures on Mass Transit and Car Use.
Sustain. (Switzerland) , 1–13, DOI: 10.3390/su12239892 (2020). Centers for Disease Control and Preventtion. COVID-19 (Coronavirus Disease). Public Health Guidance forCommunity-Related Exposure (2020).
Huang, X. et al.
Time-Series Clustering for Home Dwell Time during COVID-19: What Can We Learn from It?
ISPRS Int. J. Geo-Information , 675, DOI: 10.3390/ijgi9110675 (2020). Wang, Y., Chen, H., Ngo, V. & Luo, X. Crowdedness as the Missing Link between Shelter-In-Place and the Spreadof COVID-19, DOI: 10.2139/ssrn.3634613 (2020).
Oliver, N. et al.
Mobile phone data for informing public health actions across the COVID-19 pandemic life cycle.
Sci. Adv. , 1–7, DOI: 10.1126/sciadv.abc0764 (2020). de Montjoye, Y. A. et al. Comment: On the privacy-conscientious use of mobile phone data.
Sci. Data , 1–6, DOI:10.1038/sdata.2018.286 (2018). Siła-Nowicka, K. et al.
Analysis of human mobility patterns from GPS trajectories and contextual information.
Int.J. Geogr. Inf. Sci. , 881–906, DOI: 10.1080/13658816.2015.1100731 (2016). Vazquez-Prokopec, G. M. et al.
Using GPS Technology to Quantify Human Mobility, Dynamic Contacts andInfectious Disease Dynamics in a Resource-Poor Urban Environment.
PLoS ONE , 1–10, DOI: 10.1371/journal.pone.0058802 (2013). Glaeser, E. L., Gorback, C. S. & Redding, S. J. How much does Covid-19 increase with mobility - Evidencefrom New York and four other US cities. Tech. Rep., National Bureau of Economic Research (2020). DOI:10.3386/w27519.
Xiong, C., Hu, S., Yang, M., Luo, W. & Zhang, L. Mobile device data reveal the dynamics in a positiverelationship between human mobility and COVID-19 infections.
Proc. Natl. Acad. Sci. , 202010836, DOI:10.1073/pnas.2010836117 (2020). Goyal, P. et al.
Clinical characteristics of Covid-19 in New York city.
New Engl. J. Medicine
DOI: 10.1056/NEJMc2010419 (2020).
Wang, H. & Yamamoto, N. Using a partial differential equation with Google Mobility data to predict COVID-19 inArizona.
Math. Biosci. Eng. , 4891–4904, DOI: 10.3934/mbe.2020266 (2020). 2006.16928. Jay, J. et al.
Neighborhood income and physical distancing during the COVID-19 pandemic in the U.S. medRxiv Imbruce, V. Fostering food equity in an immigrant neighborhood of New York City during COVID-19.
J. Agric.Food Syst. Community Dev. , 1–5, DOI: 10.5304/jafscd.2020.101.028 (2020). Lauer, S. A. et al.
The incubation period of coronavirus disease 2019 (COVID-19) from publicly reported conﬁrmedcases: estimation and application.
Annals internal medicine , 577–582, DOI: 10.7326/M20-0504 (2020).
SU and RV conceived the study design. RV and TY analyzed the results. RV wrote the manuscript text. All authorsreviewed the manuscript.