key: cord-1056263-tfv8smeu
authors: Saleh, S. N.; Lehmann, C. U.; Medford, R. J.
title: Early Crowdfunding Response to the COVID-19 Pandemic
date: 2020-10-14
journal: nan
DOI: 10.1101/2020.10.12.20211532
sha: 36c60a6f9f03a34a422666d8f4bd09297a9add06
doc_id: 1056263
cord_uid: tfv8smeu

Background: As the number of COVID-19 cases increased precipitously in the US, policymakers and health officials marshalled their pandemic responses. As the economic impacts multiplied, anecdotal reports noted the increased use of online crowdfunding to defray these costs. We examined the online crowdfunding response in the early stage of the COVID-19 pandemic in the US. Methods: On May 16, 2020, we extracted all available data available on US campaigns created between January 1 and May 10, 2020 on GoFundMe and identified the subset of COVID-related campaigns using keywords relevant to the COVID-19 pandemic. We explored incidence of COVID-related campaigns by geography, by category, and over time and compared campaign characteristics to non-COVID-related campaigns after March 11 when the pandemic was declared. Results: We found that there was a substantial increase in overall GoFundMe online crowdfunding campaigns in March, largely attributable to COVID-related campaigns. However, as the COVID-19 pandemic persisted and progressed, the number of campaigns per COVID-19 cases declined more than tenfold across all states. COVID-related campaigns raised more money, had a longer narrative description, and were more likely to be shared on Facebook than other campaigns in the study period. Conclusions: Online crowdfunding appears to be a transient stopgap, predicated on the novelty of an emergency rather than the true sustained need of a community. Rather, crowdfunding activity is likely an early marker for communities in acute distress that could be used by governments and aid organizations to guide disaster relief and policy.

approval as all data were publicly available. We examined weekly new COVID-related campaigns juxtaposed with the weekly new 1 0 3 COVID-19 cases both nationally and by state or territory. We evaluated the incidence of presence of zip code level data, we also explored more granular community trends. We also 1 0 9 evaluated changes in COVID-related campaigns per million for each state and campaign 1 1 0 categories over time. We compared campaign characteristics, including fundraising, campaign information, 1 1 2 and campaign category, for COVID-19 and non-COVID-related campaigns over the same time where appropriate to determine significance. Alpha level of significance was set a priori at 0.05 1 1 5 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted October 14, 2020 . . https://doi.org/10.1101 /2020 6 and all hypothesis testing was two-sided. We did not adjust for multiple comparisons as this was 1 1 6 an exploratory study and should be interpreted as hypothesis-generating. Analyses were 1 1 7 performed using Python, version 3.7.2 (Python Software Foundation). Of 310,695 total new campaigns created between January 1 and May 10, 2020, we 1 2 0 analyzed 232,827 campaigns from the US across all categories. Twenty-two percent of these or campaigns related to or referencing COVID-19 and declined in April ( Figure 1 ). This COVID-1 2 5 related increase in campaigns was observed before the national peak of new COVID-19 cases in April. Subsequently, the incidence of new campaigns declined while the incidence of new 1 2 7 COVID-19 cases remained steady or declined slightly. Adjusting for the weekly incidence of COVID-19 cases, all US states had their peak of new COVID-related campaigns within 2 weeks of their first week when 100 new cases were 1 3 0 diagnosed in that state. For many states, the rate of COVID-related campaigns per case dropped 1 3 1 by at tenfold after the first 2 to 3 weeks ( Figure 2 ). Population rich states that also had early . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 14, 2020. . https://doi.org/10.1101/2020.10.12.20211532 doi: medRxiv preprint (n=50,828) raised a median of $930 IQR, $220-$3,075), nearly 50% more per campaign (p < a longer narrative in their campaign description, were more likely to be listed as a charity, and 1 4 7 more likely to have the campaigner be the same as the beneficiary (Table 1) .

Creators of COVID-related campaigns selected the following categories most commonly Illness & Healing peaked next within the following 2 weeks. As these initial spikes decreased is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 14, 2020. . https://doi.org/10.1101/2020.10.12.20211532 doi: medRxiv preprint 8 Early in the COVID-19 pandemic, people were more likely to create COVID-19 1 6 1 campaigns. There was a substantial increase in overall GoFundMe online crowdfunding 1 6 2 campaigns in March, mostly attributable to COVID-related campaigns. COVID-related 1 6 3 campaigns raised more money than other campaigns, had a longer narrative description, and 1 6 4 were more likely to be shared on Facebook. However, as the COVID-19 pandemic progressed, 1 6 5 its novelty wore off. As COVID-19 became a familiar ubiquitous problem, the number of chance of a campaign to "go viral", reach many people, and generate large donations may have 1 7 0 convinced potential campaign builders that COVID-related illness or financial hardships might 1 7 1 be perceived as causes that others would support. As the issue became widespread and the 1 7 2 ability to set a campaign apart as a worthy cause became more difficult, the effort (cost) of indicating a lack of a case-dependent response, supporting the hypothesis that potential campaign 1 7 8 designers perceive common illnesses as less likely to receive the attention of donors over time. While raising millions of dollars, our study suggests that online crowdfunding does not CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 14, 2020. . https://doi.org/10.1101/2020.10.12.20211532 doi: medRxiv preprint crowdfunding more likely functions as a weathervane indicating a novel crisis. Further, 1 8 4 crowdfunding is heavily swayed by marketing and storytelling as it points a majority of the funds 1 8 5 towards a few cases rather than aiding the many(8,15).

The large number of categories for COVID-related GoFundMe campaigns tells a March 2(16).

As newly generated GoFundMe campaigns showed a temporal correlation to the 1 9 6 pandemic's effect on society, we contemplated that GoFundMe campaigns may actually serve as 1 9 7 an early marker for any community in distress. To explore this hypothesis, we isolated 1 9 8 campaigns from smaller communities that have been affected by a natural disaster including 1 9 9

Puerto Rico (earthquake from January 6-7, 2020), Nashville (tornado from March 2-3, 2020), 2 0 0 and Mississippi (tornado from April 12-13, 2020). We were able to discern an immediate peak 2 0 1 for GoFundMe campaigns following each event (Appendix Figure 1 ).

While GoFundMe campaigns failed to meet the needs of those seeking support as demonstrated by the small fraction of funding goals met by data extraction date, it appears 2 0 4 however to be an accurate, early marker for a community, region, state, or country in distress. This measure is not limited to medical emergencies and extends to natural disasters. Future 2 0 6 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 14, 2020. . https://doi.org/10.1101/2020.10.12.20211532 doi: medRxiv preprint 1 0

research will show what type of negative events that affect communities may be reflected by a 2 0 7 spike in GoFundMe campaigns. We propose the development of a measure using rapid rise in 2 0 8 new GoFundMe campaigns as an early indicator of community distress.

While we were able to analyze the full available set of GoFundMe campaigns, which 2 1 0 strengthened our study, it had several limitations. First, we used data from one crowdfunding 2 1 1 platform only. While GoFundMe is the largest platform in the US by number of campaigns, our 2 1 2 study may have suffered from selection bias. Because of the large number of campaigns, we are 2 1 3 unable to validate that campaigns using COVID-related keywords were indeed designed to 2 1 4 mitigate the effect of COVID-19. Moreover, while we note significant differences in campaign 2 1 5 and funding characteristics between COVID-related and non-COVID-related campaigns, future 2 1 6 work will be important to further understand reasons for these differences and their implications 2 1 7 on long-term crowdfunding. Lastly, we only analyze the first few months of COVID-related 2 1 8 campaigns; therefore, more longitudinal analysis will be important to understand more 2 1 9 completely how peaks and troughs in COVID-19 incidence nationally and in particular, 2 2 0 communities affect campaign frequency. Online crowdfunding activity leverages a galvanized public reaction early in a public 2 2 3 health emergency like COVID-19. However, as disease spread persists and economic burden 2 2 4 continues to grow nationally, the creation of new campaigns fades. Online crowdfunding appears 2 2 5 to be a transient stopgap, predicated on the novelty of an emergency rather than the true 2 2 6 sustained need of a community. More importantly than its rather weak effect of mitigating the effect of the crisis, 2 2 8 crowdfunding activity is likely an early marker for communities in acute distress. It could be 2 2 9

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) 1 1 monitored by local, state, and federal government agencies such as Federal Emergency 2 3 0 Management Association (FEMA) as an early marker for emergencies and need for support. review board approval as all data were publicly available. Availability of data and materials: The data that support the findings of this study are available 2 3 9 upon request. Corporation. There are no other competing interests. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted October 14, 2020. . https://doi.org/10. 1101 /2020 

Narrative, median (IQR), words

Active donation days, median (IQR)

Facebook shares

Campaign Category (top 8 of 17)

https://linkinghub.elsevier.com/retrieve/pii/S1743919120303162 Comparison of Online Medical Fundraising in Canada, the United Kingdom, and the United 2 7 5 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)The copyright holder for this preprint this version posted October 14, 2020. its-very-good-for-the-market https://www.bloomberg.com/news/articles/2017-06-12/america-s-health-care-crisis-is-a-2 9 2 gold-mine-for-crowdfunding https://linkinghub.elsevier.com/retrieve/pii/S0277953617300886 2 9 7 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 14, 2020. Relief-Initiative-Small-Business-Impacted CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 14, 2020. and categorical variables as number (%). 

a Given the non-Gaussian populations, we used Mann-Whitney U, Chi-squared, and Fischer's exact testing to detect statistical differences among the three groups.. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 14, 2020 . . https://doi.org/10.1101 /2020 is defined as the first week that a state had more than 100 cases to allow for direct comparison 3 3 6 among states. All states are shown, but only the top 6 states for total COVID-19 cases per 3 3 7 million are highlighted in color (as ordered from most to least in the legend). were present before this date. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)The copyright holder for this preprint this version posted October 14, 2020. . https://doi.org/10. 1101 /2020