The contagion externality of a superspreading event: The Sturgis Motorcycle Rally and COVID-19

Abstract

Large in-person gatherings of travelers who do not socially distance are classified as the "highest risk" for COVID-19 spread by the Centers for Disease Control and Prevention (CDC). From August 7-16, 2020, nearly 500,000 motorcycle enthusiasts converged on Sturgis, South Dakota for its annual rally in an environment without mask-wearing requirements or other mitigating policies. This study is the first to explore this event's public health impacts. First, using anonymized cell phone data, we document that foot traffic at restaurants/bars, retail establishments, and entertainment venues rose substantially at event locations. Stay-at-home behavior among local residents fell. Second, using a synthetic control approach, we find that the COVID-19 case rate increased substantially in Meade County and in the state of South Dakota in the month following the Rally. Finally, using a difference-in-differences model to assess nationwide spread, we find that following the Sturgis event, counties outside of South Dakota that contributed the highest inflows of rally attendees experienced a 6.4-12.5% increase in COVID-19 cases relative to counties without inflows. Our findings highlight that local policy decisions assessing the tradeoff between local economic benefits and COVID-19 health costs will not be socially optimal in the presence of large contagion externalities.

Keywords: COVID‐19; Sturgis Motorcycle Rally; large outdoor gathering; social distancing.

FIGURE 1

Distribution of inflows into Sturgis Motorcycle Rally. (a) Absolute intensity of Sturgis attendee inflow, by Resident County. (b) Relative intensity of Sturgis attendee inflow, by Resident County [Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 2

(a) Event‐study analyses of effect of Sturgis Motorcycle Rally on nonresident smartphone pings. Estimates are generated using ordinary least squares. All models include cluster group and day fixed effects. State‐level controls include: whether food industry reopened, whether retail stores reopened, whether personal or pet care services reopened, whether entertainment or amusement activities reopened, and whether there are any mask wearing requirements. County‐level controls include: average temperature and precipitation. Confidence intervals were calculated using a wild cluster bootstrap method. (b) Event‐study analyses of effect of Sturgis Motorcycle Rally on nonresident travel and foot traffic. A, Inverse hyperbolic sine (restaurant and bar foot traffic); B, inverse hyperbolic sine (retail foot traffic); C, inverse hyperbolic sine (entertainment foot traffic); D, inverse hyperbolic sine (business services foot traffic); E, inverse hyperbolic sine (hotel foot traffic); F, inverse hyperbolic sine (campground foot traffic). Estimates are generated using ordinary least squares. All models include cluster group and day fixed effects. State‐level controls include: whether food industry reopened, whether retail stores reopened, whether personal or pet care services reopened, whether entertainment or amusement activities reopened, and whether there are any mask wearing requirements. County‐level controls include: average temperature and precipitation. Confidence intervals were calculated using a wild cluster bootstrap method [Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 3

Synthetic control estimates of effect of Sturgis Motorcycle Rally on percent stay‐at‐home full‐time. (a) Meade County. (b) Meade County and Border Counties. (c) South Dakota. The donor pool is comprised of counties with urbanicity rate ±2.5% of Meade County (62%) and weighted population density ±150% of Meade County (500), and excludes counties in border states and South Dakota, as well as counties with at least 1 home resident that pinged in Sturgis event census block groups between 8/3 and 8/16. All synthetic control estimates are generated by matching on pretreatment stay‐at‐home behavior on every day between 7/6 and 8/2 [Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 4

Synthetic control estimates of effect of Sturgis Motorcycle Rally on median hours spent at home. (a) Meade County. (b) Meade County and Border Counties. (c) South Dakota. The donor pool is comprised of counties with urbanicity rate ±2.5% of Meade County (62%) and weighted population density ±150% of Meade County (500), and excludes counties in border states and South Dakota, as well as counties with at least 1 home resident that that pinged in Sturgis event census block groups between 8/3 and 8/16. All synthetic control estimates are generated by matching on pretreatment stay‐at‐home behavior on every day between 7/6 and 8/2 [Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 5

Synthetic control estimates of effect of Sturgis Motorcycle Rally on cumulative COVID‐19 cases per 1,000 population, synthetic controls generated by matching on COVID‐19 cases on each pretreatment day. (a) Meade County. (b) Meade County and Border Counties. (c) South Dakota. The donor pool is comprised of counties with urbanicity rate ±2.5% of Meade County (62%) and weighted population density ±150% of Meade County (500), and excludes counties in border states and South Dakota, as well as counties with at least 1 home resident that that pinged in Sturgis event census block groups between 8/3 and 8/16. All synthetic control estimates are generated by matching on pretreatment stay‐at‐home behavior on every day between 7/6 and 8/2 [Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 6

Synthetic control estimates of effect of Sturgis Motorcycle Rally on cumulative COVID‐19 cases per 1,000 population, synthetic controls generated by matching on COVID‐19 cases on 14 pretreatment days, COVID‐19 testing rates, reopening policies, travel restriction policies, mask‐wearing policies and & social distancing. (a) Meade County. (b) Meade County and Border Counties. (c) South Dakota. The donor pool is comprised of counties with urbanicity rate ±2.5% of Meade County (62%) and weighted population density ±150% of Meade County (500), and excludes counties in border states and South Dakota, as well as counties with at least 1 home resident that that pinged in Sturgis event census block groups between 8/3 and 8/16. All synthetic control estimates are generated by matching on pretreatment stay‐at‐home behavior on 7/7, 7/9, 7/11, 7/13, 7/15, 7/17, 7/19, 7/21, 7/23, 7/25, 7/27/, 7/29, 7/31, and 8/2, state COVID‐19 testing rate per 100,000 population, number of days the state permitted reopening of businesses in the following sectors: restaurant, bars, movie theaters, gyms, and retail stores, and number of days the state had enacted a mask‐wearing mandate and travel quarantine mandate [Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 7

Event‐study analysis of effect of Sturgis rally on log (county‐level COVID‐19 cases), by relative inflow of county residents to Sturgis Motorcycle Rally. (a) High relative inflow; (b) moderate‐high relative inflow; (c) moderate relative inflow; (d) moderate‐low relative inflow; (e) low relative inflow [Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 8

Event‐study analysis of effect of Sturgis rally on log (county‐level COVID‐19 cases), by absolute inflow of county residents to Sturgis Motorcycle Rally. (a) High absolute inflow; (b) moderate‐high absolute inflow; (c) moderate absolute inflow; (d) moderate‐low absolute inflow; (e) low absolute inflow [Color figure can be viewed at wileyonlinelibrary.com]