Research ArticleCORONAVIRUS

Social connections with COVID-19–affected areas increase compliance with mobility restrictions

See allHide authors and affiliations

Science Advances  18 Nov 2020:
Vol. 6, no. 47, eabc3054
DOI: 10.1126/sciadv.abc3054
  • Fig. 1 Number of county-level restrictions, SCI China, and social distancing.

    (A and B) County average number of types of restrictions implemented (0 to 5) in calendar time and relative to the first restriction in the county. As of 30 March 2020, the number of state-level school closing is 52, closures of public places is 43, stay-at-home orders is 36, no-gatherings is 20, and banning nonessential services is 11. (C) Map of the SCI with China. (D) Average Social distancing plotted conditional on the county’s social connectedness with China, the first country severely affected by COVID-19. Counties are sorted into 10 portfolios by state based on SCI China. Portfolio 10 represents the highest social connectedness to China, and 1 represents the lowest.

  • Table 1 The efficacy of mobility restrictions and social connectedness.

    The dependent variable is Social distancing. Restrictions is the number of the five restriction types currently adopted in the county. SCI China and SCI Italy are the SCI values between the county and China and Italy, respectively. N cases is the current number of confirmed cases in the county. Heteroscedasticity-consistent SEs, double-clustered by county and day, are shown in parentheses. Significance levels: *0.1, **0.05, and ***0.01.

    (1)(2)(3)
    Restr. × ln(SCI China)0.0048***
    (0.0004)
    0.0011**
    (0.0005)
    Restr. × ln(SCI Italy)0.0052***
    (0.0004)
    0.0042***
    (0.0005)
    Restrictions (0–5)−0.0245***
    (0.0040)
    −0.0280***
    (0.0043)
    −0.0280***
    (0.0042)
    ln(N cases)0.0122***
    (0.0009)
    0.0114***
    (0.0009)
    0.0113***
    (0.0009)
    State-Day FEYesYesYes
    County FEYesYesYes
    N184,766184,766184,766
    R20.7560.7560.756
  • Table 2 The effect of social connectedness across risk groups.

    The dependent variable is Social distancing. Restrictions is the number of the five restriction types currently adopted in the county. SCI China and SCI Italy are the SCI values between the county and China and Italy, respectively. Diabetic is the share of county population diagnosed to have diabetes. Population density is the average population density in the county, measured in thousand people per square mile. N cases is the current number of confirmed cases in the county. Columns 1 and 2 exclude all counties with commercial airports handling at least 10,000 commercial services per year, as well as adjacent counties. Heteroscedasticity-consistent SEs, double-clustered by county and day, are shown in parentheses. Significance levels: *0.1, **0.05, and ***0.01.

    Ex. airp. & adj. countiesDiabetesPop. density
    (1)(2)(3)(4)(5)(6)
    Restr. × ln(SCI China)0.0045*** (0.0006)0.0081*** (0.0008)0.0046*** (0.0004)
    Restr. × ln(SCI Italy)0.0054*** (0.0006)0.0088*** (0.0008)0.0051*** (0.0004)
    Restr. × ln(SCI China) × Diabetic−0.0354*** (0.0057)
    Restr. × ln(SCI Italy) × Diabetic−0.0374*** (0.0057)
    Restr. × Diabetic0.1588*** (0.0341)0.1782*** (0.0360)
    Restr. × ln(SCI China) × Pop. density−0.0012*** (0.0003)
    Restr. × ln(SCI Italy) × Pop. density−0.0007*** (0.0003)
    Restr. × Pop. density0.0140*** (0.0038)0.0085*** (0.0030)
    Restrictions (0–5)−0.0159***
    (0.0058)
    −0.0200*** (0.0059)−0.0384*** (0.0057)−0.0438*** (0.0061)−0.0236*** (0.0040)−0.0273*** (0.0043)
    ln(N cases)0.0138*** (0.0020)0.0117*** (0.0019)0.0110*** (0.0009)0.0101*** (0.0009)0.0123*** (0.0009)0.0114*** (0.0009)
    State-Day FEYesYesYesYesYesYes
    County FEYesYesYesYesYesYes
    N44,40544,405184,766184,766184,766184,766
    R20.7130.7140.7570.7570.7560.756
  • Table 3 The effect of social connectedness by demography.

    The effect of social connectedness by demography.. The dependent variable is Social distancing. Restrictions is the number of the five restriction types currently adopted in the county. SCI China and SCI Italy are the SCI values between the county and China and Italy, respectively. Age is the median age of the county population. Education is the share of county population having a bachelor’s degree or higher. N cases is the current number of confirmed cases in the county. Heteroscedasticity-consistent SEs, double-clustered by county and day, are shown in parentheses. Significance levels: *0.1, **0.05, and ***0.01.

    AgeEducation
    (1)(2)(3)(4)
    Restr. × ln(SCI
    China) × ln(Age)
    0.0042***
    (0.0014)
    Restr. × ln(SCI
    Italy) × ln(Age)
    0.0041***
    (0.0014)
    Restr. × ln(Age)−0.0242**
    (0.0097)
    −0.0253**
    (0.0100)
    Restr. × ln
    (SCI China) ×
    Edu.
    −0.0037**
    (0.0017)
    Restr. × ln
    (SCI Italy) × Edu.
    −0.0018
    (0.0018)
    Restr. × Education0.0770***
    (0.0150)
    0.0624***
    (0.0155)
    Restr. × ln
    (SCI China)
    −0.0106**
    (0.0051)
    0.0044***
    (0.0006)
    Restr. × ln
    (SCI Italy)
    −0.0099*
    (0.0051)
    0.0044***
    (0.0006)
    Restrictions (0–5)0.0646*
    (0.0358)
    0.0655*
    (0.0369)
    −0.0335***
    (0.0050)
    −0.0336***
    (0.0053)
    ln(N cases)0.0125***
    (0.0009)
    0.0117***
    (0.0009)
    0.0104***
    (0.0009)
    0.0094***
    (0.0010)
    State-Day FEYesYesYesYes
    County FEYesYesYesYes
    N184,766184,766184,766184,766
    R20.7560.7560.7580.758
  • Table 4 The effect of social connectedness by political orientation and race.

    The effect of social connectedness by political orientation and race.. The dependent variable is Social distancing. Restrictions is the number of the five restriction types currently adopted in the county. SCI China and SCI Italy are the SCI values between the county and China and Italy, respectively. Republican is a dummy taking the value one if more people voted Republicans than for Democrats in the 2016 presidential election. Asian and White are the shares of county population classified as Asian and White in the Census, respectively. N cases is the current number of confirmed cases in the county. Heteroscedasticity-consistent SEs, double-clustered by county and day, are shown in parentheses. Significance levels: *0.1, **0.05, and ***0.01.

    Political orientationRace
    (1)(2)(3)(4)
    Restr. × ln(SCI China) × Rep.0.0010** (0.0004)
    Restr. × ln(SCI Italy) × Rep.0.0008* (0.0004)
    Restr. × Republican−0.0066* (0.0034)−0.0051 (0.0034)
    Restr. × ln(SCI China) × Asian−0.0601* (0.0303)
    Restr. × ln(SCI Italy) × Asian0.0723** (0.0299)
    Restr. × Asian−0.0824 (0.0896)
    Restr. × ln(SCI China) × White0.0066** (0.0032)
    Restr. × ln(SCI Italy) × White−0.0048 (0.0034)
    Restr. × White0.0060 (0.0078)
    Restr. × ln(SCI China)0.0040*** (0.0005)0.0016*** (0.0006)−0.0043 (0.0027)
    Restr. × ln(SCI Italy)0.0046*** (0.0005)0.0034*** (0.0006)0.0083*** (0.0030)
    Restrictions (0–5)−0.0199*** (0.0047)−0.0245*** (0.0050)−0.0251*** (0.0043)−0.0357*** (0.0072)
    ln(N cases)0.0129*** (0.0010)0.0120*** (0.0010)0.0106*** (0.0009)0.0122*** (0.0010)
    State-Day FEYesYesYesYes
    County FEYesYesYesYes
    N184,766184,766184,766184,766
    R20.7560.7560.7570.757
  • Table 5 Misinformation via social connections.

    Misinformation via social connections.. The dependent variable is Social distancing. Restrictions is the number of the five restriction types currently adopted in the county. SCI-weighted Low education is the SCI-weighted average of the proportion of people with less than a bachelor degree in the counties that the county itself is connected to, SCI-weighted Trump vote is the SCI-weighted average of the Trump vote share in the 2016 presidential election in the counties that the county itself is connected to, and SCI-weighted Climate denial is the SCI-weighted average of the share of people saying climate change is not happening in the counties that the county itself is connected to. All SCI-weighted variables exclude the counties located in the same state. N cases is the current number of confirmed cases in the county. Heteroscedasticity-consistent SEs, double-clustered by county and day, are shown in parentheses. Significance levels: *0.1, **0.05, and ***0.01.

    (1)(2)(3)
    Restr. × SCI-w
    Low education
    −0.2286***
    (0.0216)
    Restr. × SCI-w
    Trump vote
    −0.0010***
    (0.0001)
    Restr. × SCI-w
    Climate denial
    −0.0015***
    (0.0004)
    Restrictions (0–5)0.1664***
    (0.0162)
    0.0573***
    (0.0076)
    0.0301***
    (0.0079)
    ln(N cases)0.0185***
    (0.0011)
    0.0199***
    (0.0011)
    0.0233***
    (0.0012)
    State-Day FEYesYesYes
    County FEYesYesYes
    N185,356185,356185,356
    R20.7520.7510.749

Supplementary Materials

  • Supplementary Materials

    Social connections with COVID-19-affected areas increase compliance with mobility restrictions

    Ben Charoenwong, Alan Kwan and Vesa Pursiainen

    Download Supplement

    This PDF file includes:

    • Sections S1 to S4
    • Tables S1 to S9
    • Fig. S1

    Files in this Data Supplement:

Stay Connected to Science Advances

Navigate This Article