Skip to main content
SearchLoginLogin or Signup

Misinformation About COVID-19 and Venezuelan Migration: Trends in Twitter Conversation During a Pandemic

Published onJan 27, 2022
Misinformation About COVID-19 and Venezuelan Migration: Trends in Twitter Conversation During a Pandemic
·
history

You're viewing an older Release (#5) of this Pub.

  • This Release (#5) was created on Apr 10, 2022 ()
  • The latest Release (#6) was created on May 24, 2022 ().

Abstract

This article asks whether and how—during the COVID-19 pandemic at a time of considerable uncertainty—current events and announcements by governments and political leaders are associated with trends in Twitter conversation. Using Spanish-language tweets, we examine the changing dynamics of misinformation conversation about the COVID-19 virus, international border closures, and the sociopolitical reception of Venezuelan migrants returning home during the pandemic amid an ongoing refugee crisis. We identify specific events and statements made by governments and political leaders and assess how they relate to trends in conversation about COVID-19 and migration misinformation. Findings from an analysis of time-series data reveal that several specific announcements are associated with structural shifts in Twitter conversation, and that shifts in misinformation trajectories can take different forms.

Keywords: COVID-19, displacement, migration, misinformation, Venezuela


1. Introduction

The global rise of misinformation has powerful consequences for politics, economies, and public health (Scheufele & Krause, 2019). False facts are especially dangerous on social media platforms, where ideas and information can rapidly spread unchecked. Rumors can cascade in orders of magnitude greater than factual information, especially when inaccurate facts are novel or emotionally charged (Vosoughi et al., 2018). This swift spread is further amplified by retweets from bots and highly connected users functioning as information hubs within the social media network (Bovet & Makse, 2019; Shao, Ciampaglia et al., 2018; Shao, Hui et al., 2018), and by political leaders who are well-positioned to spread misinformation in all types of media outlets (Maurer & Reinemann, 2006; Van Duyn & Collier, 2019).

Misinformation sways election outcomes, alters local and national responses to climate change, and garners support for war (Jackson & Jamieson, 2007). Fake news producers generated significant amounts of misinformation during the 2016 U.S. presidential election, particularly about Hillary Clinton, possibly influencing the outcome of the election (Bode et al., 2020). During the 2020 U.S. presidential debates, political elites’ distortions of the truth facilitated an increase in misinformation discussion on Twitter, in newspapers, and on cable news for partisan gain (Haber et al., 2021). Misinformation also threatens public health, from inaccurate advertisements about mouthwash that causes cavities to false information about the transmission and treatment of contagious diseases such as Ebola and COVID-19 (Jackson & Jamieson, 2007; Oyeyemi et al., 2014; Ricard & Medeiros, 2020). Biased or incomplete framing of immigrants erodes public support for, and the passage of, immigration policy reform, and misinformation in social media contributes to the growth of anti-immigrant and racist sentiment (Ekman, 2019; Haynes et al., 2016). Misinformation also targets refugees and asylum seekers as they assess the risks of migrating versus staying in their home country and weigh their options to find safety (Gerver, 2017; Ruokolainen & Widén, 2020).

In this article, we offer new insights about the dynamics of Spanish-language Twitter conversations about misinformation and specific events and announcements that influence that conversation. Amid a global pandemic and disruptions to international migration, understanding conversation dynamics in languages beyond English and in multiple geopolitical contexts is more pressing than ever. We focus on public health and migration misinformation, which are especially salient and dangerous in the context of a pandemic. We examine the spread of false information about the COVID-19 virus, international border closures, and the sociopolitical reception of Venezuelan migrants returning home during the pandemic amid an ongoing refugee crisis.

At a time of considerable uncertainty, we focus on whether and how trends in Twitter conversations about migration misinformation vary for three sets of events. The three illustrate distinct dynamics at the center of the pandemic: a major global public health announcement, an announcement about a false treatment for COVID-19 by a major political leader, and a set of announcements about entry and exit restrictions at the Venezuelan and Colombia borders. Using Spanish-language tweets shared on Twitter, we ask whether and how these events and announcements shape online conversation about COVID-19 and migration misinformation. Findings offer new insights about the dynamics of false information in conversation on Twitter, where few monitoring systems exist. Shifts in online conversations occur after the three types of events, although conversation trajectories take different forms.

2. Public Health and Migration-Related Misinformation

Many studies examine public health misinformation on social media platforms. The most widely examined topics relate to vaccination, Ebola, and the Zika virus, followed by cancer, smoking, and nutrition (Y. Wang et al., 2019). Overall, studies show that public health misinformation is spread widely on social media platforms. For example, most tweets about Ebola during a 2014 outbreak in Guinea, Liberia, and Nigeria contained misinformation and had further reach than tweets containing accurate information (Oyeyemi et al., 2014). In addition, misleading Facebook posts about the Zika virus were much more popular than posts with accurate public health information (Sharma et al., 2017).1 Vaccine myths are a key topic on social media and vaccine misinformation is common. Dredze et al. (2016) show that pseudo-scientific claims about a Zika vaccine rose rapidly as general interest in vaccine development for the Zika virus grew, suggesting that consumers quickly accepted false claims as true. Guidry et al. (2015) examine how vaccines were portrayed on Pinterest; most content was anti-vaccine or fed conspiracy theories about vaccine development, as people expressed concerns about safety and side effects.

In addition to public health misinformation, studies about social media conversations also examine content with respect to hate speech, racism, and xenophobia. Research shows that Twitter data offer insights about the use of different kinds of racist and anti-immigrant terms and the context in which they are deployed (Chaudhry, 2015). Kreis (2017) and Erdogan-Ozturk and Isik-Guler (2020) find that anti-refugee discourse on Twitter includes narratives about refugees as criminals, economic threats, invaders, and out-group members, emphasizing race-based notions of nationalism. These studies focus on conversation in a single hashtag rather than broader conversation expressed on the platform. For example, Erdogan-Ozturk and Isik-Guler (2020) analyze refugee discourse express in the hashtag, #IDontWantSyriansInMyCountry and report that Twitter users construct refugees as a dangerous out-group and simultaneously form a collective Turkish in-group identity. Other studies analyze a random sample of tweets to examine anti-refugee discourse and sentiment. Bozdağ (2019) shows that Twitter users responded to national debates about granting citizenship to Syrian refugees in Turkey by tweeting about refugees as economic, cultural, and social threats. The differences in methodological approaches to identifying false and xenophobic tweets illustrate the complexity of studying migration misinformation on social media platforms. As Kreis (2017) notes, because xenophobic and exclusionary rhetoric may emerge through direct and indirect language, it poses challenges to grasping the complexity of racist expressions in digital discourse, such as how to bridge the gap between micro/individual and macro/social structural levels (Ozduzen et al., 2020).

3. Current Events, Political Announcements, and the Spread of Misinformation

The central interest in this article is how—during the COVID-19 pandemic, a period of great uncertainty—current events and announcements by political leaders and governments relate to online conversations. Specifically, are specific announcements or events associated with trends in Twitter conversation about COVID-19 and, if so, how?

Prior studies on gun violence illustrate one dynamic. For example, after the 2012 shooting at Sandy Hook Elementary School in Connecticut, N. Wang et al. (2016) observed a spike in tweets about gun control and, despite elevated rates of both pro-gun and anti-gun sentiments, pro-gun sentiments persisted longer in online conversations. In the wake of the shooting, Benton et al. (2016) showed that current events and developments in gun control legislation were associated with spikes in the number of tweets.

Statements by political leaders also influence conversations on Twitter (Maurer & Reinemann, 2006; Van Duyn & Collier, 2019). In a study of statements proffered by then-presidential candidate Donald Trump in the 2016 U.S. presidential election, Hernandez (2018) documented a rise in online conversations and in specific hashtags such as migrants being “bad hombres.” After Trump made such statements, Hernandez (2018) found it took only 14 min for memes to emerge on Twitter. These dynamics also play out beyond the United States. After Turkish president Erdogan announced that he would grant citizenship to Syrian refugees, Bozdağ (2019) found that Twitter users in Turkey ignited a new wave of xenophobic and anti-refugee tweets that provided justification for discriminatory narratives about Syrian refugees by citing false information and reinforcing an ‘us vs. them’ discourse.

Although misinformation as promoted by political elites is well established in prior studies, its spread in languages other than English on social media platforms and in state-run media is not clearly understood (Hernández-Garcia & Giménez-Júlvez, 2020). Some scholars are developing methodologies to detect deceptive information in Spanish and other non-English languages, but few papers have reported substantive findings on specific patterns or types of misinformation that gain traction in Spanish (Posadas-Durán et al., 2020; Pranesh et al., 2021). Furthermore, most studies about elites and social media focus on misinformation in countries with free presses. Thus, it is also unclear how these dynamics are amplified or exacerbated in state-run media environments.

One exception is Russia, where the government spreads digital propaganda through trolls and bots2 to manage coverage of specific events such as the 2013 Ukrainian conflict (Aro, 2016; Sanovich, 2017). Although the misinformation promulgated by Russian president Vladimir Putin and former president and prime minister Dmitry Medvedev usually focuses on domestic issues, Bolgov et al. (2019) analyze whether and how their Twitter accounts contribute to the spread of misinformation about global issues. In addition, the Chinese government prioritizes traditional state-run news platforms to perpetuate propaganda and misinformation (Guo, 2020). Recent efforts to promote Chinese Communist Party propaganda on Twitter is spread by a small number of Twitter accounts managed by mid-level diplomats, rather than high-level party leaders (Huang & Wang, 2019). Thus, it remains unclear to what extent state-run media influences misinformation conversation on social media platforms. In this article, we focus on misinformation discussion related to Venezuela, a country in which President Nicolás Maduro and his national government runs and controls media messaging by censoring independent journalism and limiting internet freedom.

4. Key Events in This Study

This article empirically examines whether and how specific events and political announcements shift the underlying structure of Twitter conversation about COVID-19 and migration. The first component of our analysis examines vaccine misinformation conversation as it relates to the December 11, 2020, announcement that the Pfizer-BioNTech vaccine was approved for emergency distribution in the United States (U.S. Food and Drug Administration, 2021). Pfizer was the first pharmaceutical company to seek approval in the United States and United Kingdom, after releasing results from its clinical trial (AJMC, 2021). In the weeks that followed, Moderna and AstraZeneca also filed for approval to distribute their COVID-19 vaccines. We focus on the Pfizer announcement because it was the first vaccine approval.3 In this case, we consider the impact of this announcement followed by a series of similar announcement events.

In a second analysis, we consider a specific case of COVID-19 misinformation by an elected official, Venezuelan president Nicolás Maduro. On January 24, 2021, he declared that Venezuelan scientists discovered “miracle droplets” that cured COVID-19 with “100% efficacy” (Associated Press, 2021). He tweeted multiple times about this “miracle cure” during January 2021, and shared videos on Facebook ostensibly showing the laboratory development of the “miracle droplets.” In response, Facebook banned him for 30 days because of this announcement, citing it as misinformation that threatened public health (Ellsworth, 2021). In this case, we study the impact of a single announcement by a prominent leader.

In the third analysis, we focus on a set of contradictory migration policies and announcements made by Venezuelan president Maduro during the summer of 2020. Because of economic collapse, resource shortages, and political turmoil, since 2015 more than five million Venezuelans have left the country seeking protection and economic opportunity elsewhere (Duddy, 2015; United Nations High Commissioner for Refugees, n.d.; Viscidi, 2016). With the pandemic’s onset, economic shutdowns and border closures throughout the region further amplified the dire conditions that many displaced Venezuelans faced. As a result, more than 100,000 Venezuelans returned to Venezuela from Colombia alone, despite border closures and persistent political and economic challenges (Méndez-Triviño, 2020). Early on in the pandemic, Maduro and the Venezuelan government sent mixed and contradictory messages about return migrants. On June 8, 2020, they imposed border restrictions at the Simón Bolívar bridge, the largest border crossing point between Colombia and Venezuela (International Crisis Group, 2020). Three days later, on June 11, Maduro announced he welcomed Venezuelan citizens living abroad back into the country, promising to treat migrants “equally” with “love” and “dignity” (VTV Canal 8, 2020). In mid-July, the Venezuelan military launched a campaign against migrants returning through unauthorized channels, denouncing them as “bioterrorists” carrying “biological weapons” to infect Venezuelans (Associated Press, 2020; Pardo, 2020). During this time, human rights organizations reported that return migrants were held in unsanitary quarantine detention facilities for more than a month, exceeding the World Health Organization’s (WHO) 14-day quarantine protocol (Human Rights Watch, 2020; Kurmanaev et al., 2020). In this example, we examine the impact of several contradictory announcements and policies during a short time frame.

These three analyses capture misinformation-related conversation during the COVID-19 pandemic at international and national levels. The Pfizer vaccine announcement offers global insights about misinformation conversation as the pandemic’s course shifted with the promise of a vaccine. The second and third analyses reveal dynamics specific to Venezuela, the largest migrant-sending country in South America (International Organization for Migration, 2021). Maduro’s “miracle cure” announcement provides a window into one specific case of misinformation spread by an elected official in the context of state-run media. Subsequent focus on contradictory migration policies in the third analysis suggests how the COVID-19 pandemic can exacerbate long-standing geopolitical crises. To observe the change in volume of misinformation conversation during the pandemic, our analysis spans the period of March 2020 to March 2021 and focuses on specific time windows related to each selected event/announcement. In other words, do these events lead to a reduction or an increase in conversation about the misinformation topic? 

We pause to mention that disentangling support or rebuke of specific misinformation is beyond the scope of this article. Instead, our focus is on modeling the dynamics of the volume of discussion about these misinformation topics and understanding how those dynamics change (or not) after specific announcements. Emphasizing the volume of misinformation discussion is important given psychological studies that document the ‘illusionary truth effect,’ for example, how beliefs derive from familiarity rather than objective truth (Fazio et al., 2015; Hasher et al., 1977; Henkel & Mattson, 2011; Pennycook et al., 2018; Stanley et al., 2019). Therefore, if people hear the same misinformation many times, their beliefs tend to become long-lasting as the misinformation appears to be more true than false (Lewandowsky et al., 2020).

5. Data

We rely on two data streams from Twitter to observe trends in misinformation conversation about COVID-19 and migration during the pandemic. First, we analyze tweets from the Twitter Decahose stream, a 10% random sample of all daily tweets4 (Decahose application programming interface [API]) for the period August 1, 2020, through January 31, 2021. From the subset that Twitter labels Spanish-language, we created two conversation streams or subgroups of tweets, one related to COVID-19 and another related to migration. The COVID-19 stream consists of tweets that contain pandemic-related words such as ‘COVID,’ ‘coronavirus,’ and ‘pandemic.’ The migration stream includes tweets that contain words such as ‘asylum,’ ‘refugee,’ and ‘migrating.’

Second, we collect tweets using the Twitter Keyword/Hashtag API to collect tweets containing specific hashtags or keywords. Since January 15, 2020, we have collected tweets that contain COVID-19 related hashtags in English. Because of the universality of hashtags such as #covid and #coronavirus, this data set contains a large number of non-English tweets. When using this data stream, we focus on tweets containing one of the COVID-19 related hashtags and Spanish-language terms capturing two specific events, one in June 2020 and one in January 2021. Although there are different preprocessing strategies to enhance the quality of topic identification, we follow Churchill and Singh (2021) to create our topic content analysis. Preprocessing steps include normalizing capitalization; removing punctuation and URLs; removing stop-words (such as content-poor words ‘and’ and ‘the’); and word stemming (to adjust words for grammatical variants in conjugation, and gender and plurals in Spanish).

5.1. COVID-19 Misinformation Conversation

We begin by examining how conversation about specific myths related to COVID-19 gain traction on social media. To determine what myths to include in the analysis, we identified the most common, salient, and dangerous Spanish-language myths defined by the WHO and discussed in media coverage of misinformation in Latin America (Álvarez, 2020; Armus, 2021; Fisher, 2020; Frenkel, 2021; Gardel, 2020; Lodoño & Simões, 2020; Lodoño et al., 2021; Martínez, 2020; World Health Organization, 2020; World Health Organization, 2022). Beginning with a list of frequently occurring Spanish words and two-word phrases, and manually identifying other phrases capturing WHO public health messaging, domain experts hand-curated a dictionary of words and phrases for 43 specific myths. We then used these preprocessed dictionaries to identify tweets containing any of the myths in our myth dictionaries, which capture core vocabulary related to each myth. We use the dictionaries to determine whether a post is discussing a specific myth, rather than evaluate the stance or position of users posting the myth.

We then constructed a frequency variable representing the daily volume of tweets discussing each myth. From there, we manually chose the final list of myths based on how often they appeared and how dangerous they were. This list contained 20 myths ranging from conspiracy theories about the origin of the virus to home remedies that would cure the disease. From these, we identified four high-level themes of misinformation conversation: COVID-19 origins; COVID-19 transmission; prevention, treatments, and cures; and vaccines. Appendix Table A1 shows the translated original list of 43 myths and the final list of 20 myths (indicated by an asterisk). It also presents the mapping of the myths to each misinformation theme. In the analysis that follows, we focus only on conversation related to vaccine myths as one form of misinformation conversation taking place during the pandemic.5 To examine how the volume of vaccine myth conversation shifts across time, we use tweets from August 2020 through January 2021. This period captures when conversation about vaccine approval first began, starting with a Russian pharmaceutical company registering the Sputnik V vaccine before finishing clinical trials through Pfizer-BioNTech vaccine approval and rollout in late 2020 and early 2021, and subsequent announcements by other pharmaceutical companies.

5.2. COVID-19 Misinformation and the President of Venezuela

The second part of our analysis involves a deep dive into conversation about a supposed “miracle cure” endorsed by Venezuelan president Maduro. From newspaper coverage of Venezuela between January 2020 through March 2021, we identified key events related to COVID-19 misinformation generated and spread by Maduro and his government. In the analysis, we track volume shifts in Twitter conversation and their association with President Maduro’s proclamation that Venezuelan scientists found a miracle cure for COVID-19. Once again, we employed a similar approach to that used to examine COVID-19 misinformation conversation. Subject-matter experts hand-curated a dictionary of Spanish words and phrases based on preliminary research and the frequency of Spanish words and phrases on Twitter referring to the miracle cure. This dictionary contained two key myths related to Maduro’s announcement, and they averaged four words and phrases per myth. Using the dictionary, we identified relevant tweets, calculated daily frequencies, and grouped specific myths into the miracle cure theme. Appendix Table A1 contains the myths and examples of translated words and phrases associated with each myth. The period for this analysis covers 45 days before and after Maduro’s announcement about the discovery of such a cure, from December 10, 2020, through March 10, 2021.

5.3. Migration Misinformation

The third, and final, component of the analysis examines trends in the misinformation conversation about migration, border crossings, and Venezuelan displacement. Because the pandemic closed international borders and interrupted global migration patterns, especially for refugees and asylum seekers, we expect to see conversation about forced migration as well as misinformation about migration and COVID-19. Drawing on prior migration studies and analyzing Spanish-language media coverage of Venezuela and the surrounding region from January 2020 to March 2021, we identified a series of contradictory statements about migration and border policies made by the Venezuelan government in June and July 2020.

Similar to the earlier approach, we generated a hand-curated dictionary containing Spanish words and phrases that captured these contradictory statements, based on media coverage and a preliminary analysis of frequent words and two-word phrases in Spanish on Twitter. We then used the dictionary to identify relevant tweets, computed the daily frequency of each theme, and examined how the themes shifted over time. We generated dictionaries for three high-level topics: border restrictions, “Venezuelans are welcome back,” and “migrants are bioterrorists.”6 This analysis focuses on trends in Twitter conversation about border restrictions to gauge whether and how the conversation shifts given Maduro’s contradictory announcements. We used tweets from March 2020 to October 2020 to capture a four-month window before and after the Venezuelan government reduced its daily quota for migrants entering the country in June 2020, and then examined trends in the frequency of misinformation conversation about border restrictions and migration. Notably, this window also captures other key migration events in the summer, including Maduro’s speech welcoming Venezuelan migrants back home and, a few weeks later, a military campaign that launched labeling return migrants as “biological weapons.” 

5.4. Data Dictionary Creation

Table 1. Summary of themes and topics

Misinformation type

Misinformation theme

Example topic phrases

COVID-19 Misinformation

Vaccine myths

Vaccine has a microchip; vaccine side effects are worse than COVID; Vaccine alters DNA

Venezuelan COVID-19 misinformation 

Miracle cure

Miracle droplets, miracle cure, Carvativir

Venezuelan migration misinformation

Border restrictions

Open border, illegal crossing, Táchira

Table 1 summarizes the misinformation themes and topics related to COVID-19 and migration that we analyze in this article.7 The vaccine myths theme captures incorrect ideas that include: the vaccine includes a microchip, its side effects are worse than COVID-19 itself, and it can alter someone’s DNA. The miracle cure theme includes mentions of miracle droplets, miracle cure, and Carvativir, for example, the chemical base of the supposed cure. Finally, Table 1 presents the border restrictions theme, which includes myths related to open borders, illegal crossings, and Táchira, a Venezuelan/Colombian border crossing station that the Venezuelan government often targets in its rhetoric about returning migrants. (For more details, see Appendix Table A1.)

6. Analytic Strategy

For each section of the analysis that follows, we begin by describing how specific announcements or events align with trends in misinformation conversation on Twitter. We plot the time-series data and mark dates of key announcements to visually display trends in Twitter conversation during the specified period. We follow by using the autoregressive integrated moving average (ARIMA) modeling framework (Box & Jenkins, 1970) to assess whether and how specific announcements are associated with the volume of Twitter misinformation conversation about vaccines, a COVID-19 cure, and border restrictions. ARIMA modeling is commonly used for (1) descriptive analysis to understand the structure of specific time series and (2) for forecasting purposes. In this work, we use ARIMA modeling to understand the structure of different time series, permitting us to identify the model that best fits the time-series data before and after each announcement or event. We can then use the model parameters to determine whether an event or series of events is associated with structural shifts in Twitter conversation.8 For our purposes, the ARIMA models offer insights about how announcements are related to social media conversations about misinformation, specifically whether these shifts are temporary bursts or a longer-term restructuring of the observed data. To this end, we use ARIMA modeling to understand the nature of differences between and pre- and post-event time series and to identify if the models that best describe the pre- and post-event time series are structurally different (such that the AR and MA components shift significantly—see below) or if the models have similar components, suggesting that there is a short-term burst in volume and then an eventual return to the original time-series structure. If the models are structurally different, we can conclude that the event(s) are contributing to a longer-term restructuring of the data.

We develop the technical specifications of the model in four steps. First, we examine the overall time series in a standard plot and observe notable trends. Second, after making an initial visual assessment, we estimate the parameters of the transfer function using autocorrelation functions (ACF) and partial autocorrelation functions (PACF) of the data as well as the differenced data to determine which type of model and lag interval to employ. We measure the common trends that help identify the possible structure for an ARIMA model, such as autoregressive terms, moving average terms, and the appropriate number of lags used to confirm the selected models.

Third, we estimate an ARIMA model for the pre- and post-event periods. We use an unconstrained Hyndman-Khandakar algorithm for automatic ARIMA modeling to fit the best model to the data before and after the event of interest, and conduct an exhaustive search of the set of possible models by adjusting stepwise and approximation parameters (Hyndman & Khandakar, 2008). We select the best models based on considering both the Akaike’s information criteria (AIC) and Bayesian information criteria (BIC), as well as root mean squared error (RMSE). The analysis is based on the general form ARIMA(p, d, q) models, where p is the number of autoregressive terms, d is the number of nonseasonal differences needed for stationarity, and q is the number of lagged forecast errors (MA terms). Let {yt{y_t }} denote the time series of Twitter data, and dyt∇^d y_t denotes the differencing at d levels.9 Then for the case with d=1d=1 (first differencing), we have

yt=α0+α1yt1++αpytp+wt+β1wt1++βqwtq,∇y_t=α_0+α_1 ∇y_{t-1}+⋯+α_p ∇y_{t-p}+w_t+β_1 w_{t-1}+⋯+β_q w_{t-q},

where yt∇y_t denotes the differenced value of yty_t with one level of differencing, and {wtw_t} denotes a zero-mean Gaussian white noise process with constant variance. 

Fourth, we run two analyses to check that the models appropriately fit the data. We accomplish this by generating residual plots based on the ACF and PACF of the model residuals to confirm that the models provide adequate fits to the data. We then use measures of fit to compare whether the post-event models better fit the post-event data than the pre-event models, and use AIC, BIC, and RMSE to evaluate whether estimates of the pre- and post-event data represent a structural change related to key events in the time series. The best models have the lowest values for RMSE, AIC, and BIC.

Finally, we assess the structural differences between the best models for pre- and post-event data. We conclude a structural shift has occurred if the best models for pre- and post-event data are structurally different, that is, if models indicate difference in stationarity (stationary and nonstationary behavior) or if they are different in AR and MA components as these components describe different autocorrelation behavior. If the event(s) is not significant, we would expect that the best models for pre- and post-event data to be similar in structure.

We note that the dates of announcements or events in time-series analysis play an important role in the construction of ARIMA models, such that inaccurate or ambiguous dates may lead to a broad range of possible models (Gilmour et al., 2006). Drawing on the utility of both data-driven and social contextual approaches, we take up Isaac and Griffin’s (1997) call for historical “actuality” in social scientific research and selected events that are historically grounded and theoretically informed. We identified events related to COVID-19 and migration since the start of the pandemic by analyzing news reports from a variety of Spanish- and English-language media sources.

7. Findings

7.1. COVID-19 Misinformation

Figure 1. Time-series plot for vaccine myths. From COVID Decahose of Spanish tweets (August 2020–January 2021).

Figure 1 presents the volume of conversation about vaccine myths from August 2020 through January 2021, when Twitter conversation about vaccine approval first began. The three vertical lines refer to the day and year of the registration of the Russian Sputnik V vaccine, President Maduro’s announcement that the Sputnik V vaccine arrived in Venezuela, and the U.S. government’s approval of the Pfizer vaccine. Twitter conversation about vaccine myths spiked immediately after the Sputnik V vaccine announcement, although it was not sustained. Thereafter, the overall volume of vaccine misinformation conversation was relatively low until November 2020, when the Pfizer, Moderna, and Astra-Zeneca companies began publishing results about vaccine efficacy. Although a small rise in conversation about vaccine myths follows, the misinformation conversation picks up in December when the U.S. government approved the Pfizer vaccine and its rollout. These shifts, especially the linear upward trend after the Pfizer announcement, suggests that the time series is a nonstationary process.

Figure 2. Plot of ARIMA Models before and after vaccine approval. From COVID Decahose of Spanish tweets (October 2020–January 2021).

To conduct the ARIMA analysis, we split the data into pre- and post-approval of the Pfizer vaccine. The Phase 1 model includes Twitter conversation about vaccine myths between October 10 and December 10, 2020, whereas the Phase 2 model includes Twitter conversation about vaccine myths from December 11, 2020, to January 31, 2021. Appendix Table A2a presents the parameter estimates of the ARIMA model. The findings suggest there is a structural change in Twitter conversation about vaccine myths before and after the Pfizer announcement. The pre-event model is best fitted by an ARIMA(5,1,0), and the post-event model is best fitted with an ARIMA(0,0,1). The pre-event model is nonstationary with five AR lags and the post-event model is stationary with one MA lag, indicating substantially different behavior between the two sets. Figure 2 displays the observed values from Figure 1 and the fitted values derived from the ARIMA modeling. Before the Pfizer announcement, we can see—at best—small peaks between November and December 2020. After December 11, 2020, the fitted values reveal a larger volume of vaccine myth conversation that contains several upward spikes. To investigate whether the ARIMA models adequately account for the autocorrelation in the data and are an appropriate fit, we examine residual plots and confirm that residuals do not show any significant autocorrelation based on 95% confidence bands, indicating that the models adequately capture the autocorrelation in the data (residual plots available upon request).

Finally, we calculate measures of fit to evaluate whether the post-event model outperforms the pre-event model using our three criteria. Appendix Table A2b reports the goodness of fit measures that suggest the best pre- and post-event models have similar levels of fit and error. The RMSE is lower for the post-event versus pre-event model. The RMSE for the forecast errors using the pre-event model to forecast the post-event data is much larger than the RMSEs for both the pre-event model (for the pre-event data) and post-event model (for the post-event data). This indicates the lack of fit of the pre-event model to the post event data. Thus, we conclude that the best models for pre/post events are substantially different and a structural shift has occurred.

7.2. COVID-19 Misinformation and the Venezuelan President

Figure 3. Time-series plot for ‘miracle cure.’ Maduro refers to President Nicolás Maduro of Venezuela. From COVID hashtag stream (December 2020– March 2021).

The second component of the analysis examines trends in miracle cure misinformation conversation and related announcements made by Venezuelan President Maduro. Figure 3 shows that conversation about this type of miracle cure misinformation spiked immediately after Maduro’s three announcements in January 2021. The spike after the announcement that the cure arrived is especially large compared to spikes for the other two announcements. Yet, the conversation disappears within weeks of each announcement.

Figure 4. Plot of ARIMA Models before and after Maduro’s “miracle cure” announcement. ARIMA = autoregressive integrated moving average. From COVID hashtag stream (December 2020–March 2021).

To estimate the ARIMA model, we split the data into pre- and post-announcement about the arrival of a miracle cure. The Phase 1 model includes misinformation in Twitter conversation between December 10, 2020, and January 24, 2021; the Phase 2 model includes miracle cure misinformation from January 25, 2021, to March 10, 2021. Appendix Table A3a presents the parameter estimates of the ARIMA model. Figure 4 displays both the observed values and fitted values from the ARIMA modeling. Once again, the modeling suggests there is a structural change in Twitter conversation about a COVID-19 miracle cure before and after Maduro’s announcement. The pre-event model is best fitted by an ARIMA(0,0,1), and the post-event model is best fitted with an ARIMA(4,1,0). Before the announcement that a miracle cure arrived, we see only a few small peaks in Twitter conversation. Afterward, however, although the fitted values reveal a larger volume of miracle cure misinformation conversation, the upward trajectory is sustained only for several weeks.

Similar to the analysis for COVID-19 misinformation, we examine model adequacy by plotting the residuals and calculating measures of fit. The ACF and PACF residual plots indicate that pre- and post-event models adequately capture the autocorrelation present in the data, as the values consistently fall within the 95% confidence intervals. Appendix Table A3b shows the measures of fit calculations. They offer evidence of a structural change in Twitter conversation before and after Maduro’s announcement about the miracle cure. The RMSE, AIC, and BIC values for the post-event model are lower, although the best pre/post models have similar levels of fit and error.10 The RMSE for the forecast errors using the pre-event model to forecast the post-event data is substantially larger than the RMSEs of both pre-event and post-event models. Thus, we conclude that the best models for pre/post events are substantially different and a structural shift has occurred.

7.3 Migration Misinformation

Figure 5. Time-series plot for Venezuelan border restrictions. From COVID hashtag stream (March 2020–October 2020).

Figure 5 describes a rise in Twitter conversation about border restrictions during the weeks following March 17, 2020, when Venezuela closed its international borders in response to the COVID-19 pandemic. In contrast, in June and July 2020 when the government issued contradictory announcements that tightened border restrictions, welcomed Venezuelan migrants back, and launched a campaign against migrants as bioterrorists, we see small spikes in the conversation about border restrictions.

Figure 6. Plot of ARIMA Models before and after heightened border restrictions. ARIMA = autoregressive integrated moving average. From COVID hashtag stream (March 2020–October 2020).

The parameter estimates from the pre-event and post-event ARIMA modeling (see Appendix Table A4a) suggest a structural change in the data before and after Venezuela imposed new limits on daily border crossings in June 2020. The pre-event model is set to an ARIMA(4,1,0) whereas the post-event model is best fitted with an ARIMA(1,1,1), indicating a structural shift in the conversation about border restrictions with the introduction of contradictory policies. Figure 6 presents the observed and fitted values and reveals a structural shift in migration misinformation conversation before and after the announcement of restrictions at the Venezuelan border.

We ran checks on how well the pre- and post-event models fit the data. The ACF and PACF residual plots confirm that both pre- and post-event models adequately capture the autocorrelations in the data. Appendix Table A4b reports the goodness of fit measures. Once again, the goodness of fit measures suggest that the post-event model somewhat outperforms the pre-event model; the AIC, BIC, and RMSE are all lower in the post-event model and model differences in fit and error are small.11 The RMSE for the forecast errors using the pre-event model to forecast the post-event model is much larger than the RMSEs of both pre-event and post-event models. Thus, we see evidence of a structural change in conversation about border restrictions before and after the contradictory policies were implemented.

8. Discussion

This article examines how—during the COVID-19 pandemic, a period of great uncertainty—current events and announcements by government and political leaders relate to online conversations. We identified specific events and announcements that capture key dynamics of the pandemic (vaccine rollout; misinformation about COVID-19 treatment and cure; and border restrictions and closures), and then examined whether and how they associate with trends in misinformation about COVID-19 and migration in Twitter conversations. To expand research on misinformation, we analyzed social media conversation in Spanish-language tweets in Venezuela, a state-run media environment. Amid a global pandemic and global disruptions to migration, understanding the prevalence of misinformation in languages other than English and in multiple geopolitical contexts is more pressing than ever.

We first examined vaccine myths and asked whether and how conversation volume about such myths shifted before and after the announcement of the Pfizer-BioNTech vaccine approval and rollout in December 2020. Compared to before the Pfizer announcement, afterward we see a larger volume of vaccine myth conversation with several upward spikes. In addition, findings from ARIMA modeling revealed that the best pre- and post-event models are not the same, suggesting that the observed pre/post shift toward a persistent upward trajectory was structural. We also examined trends in miracle cure misinformation conversation volume after announcements made by Venezuelan president Maduro. Here we see a burst of misinformation immediately after the “miracle cure” announcement in January 2021, but the spike was short-lived and disappeared within a few weeks. Results from the ARIMA modeling also revealed there was a structural change in Twitter conversation before and after this announcement. Finally, we focused on trends in Twitter conversation volume about border restrictions starting in March 2020, when Venezuela closed its international borders in response to the COVID-19 pandemic, through summer 2020, when the government issued a series of contradictory policies and statements about migrants returning to the country. Once again, findings from the modeling reveal evidence of a structural change in conversation volume about border restrictions before and after other contradictory policies were implemented. In this case, the structural shift is associated with a sustained conversation at a lower level than before the series of announcements.

Together, the findings suggest that different types of events can create different trends in the distribution of misinformation conversation about COVID-19 and migration. This idea is consistent with prior studies that document differences in the ways in which political elites influence the spread of misinformation (Haber et al., 2021; Maurer & Reinemann, 2006; Van Duyn & Collier, 2019). Prior studies also suggest that differences in the diffusion of misinformation conversations also reflect the actions of social media platforms. For example, when Facebook banned President Maduro from its platform saying his “miracle cure” announcement was misinformation, the action likely affected the misinformation conversation on other social media platforms.

These findings suggest several avenues for future research. For example, although we performed a standard time-series analysis that identified how specific announcements relate to misinformation conversation volume, future research could employ event-detection analysis to simultaneously understand multiple events and time series. Examining these interactions and how they shift (strengthen or weaken) with other factors is an important next step. How do simultaneous conversations about vaccine and miracle cure misinformation feed off each other, and then affect discussions about one versus the other? In addition, building on preliminary analysis presented in footnote 10, future research could integrate data from other sources, including newspapers, and correlate coverage in these sources (especially in state-sponsored news discussions prevalent in Venezuela) with speeches by political leaders and online social media conversation. Future research could also identify how much of the observed misinformation was debunking vs. actual misinformation, consider interventions designed to stop the unpoliced spread of such information, especially in non-English languages, and examine change in the sentiment and stance in conversation before and after an intervention.

On the whole, the findings suggest that misinformation conversation correlates with external announcements and events, which then leads to greater exposure of false information among vulnerable populations. The exposure may take different forms, including sustained growth in misinformation conversation, a short-term upward burst that quickly disappears, or a decline that then maintains itself over time. Thus, whether and how different types of events influence online conversation and its consequences are critically important questions to address. Although online conversation has great potential as a data source to answer many types of research questions, we caution that it also involves misinformation and future research must consider these shifts in conversation dynamics.


Disclosure Statement:

We are grateful to Georgetown University’s Institute for the Study of International Migration, Georgetown University’s Massive Data Institute, MDI Impacts, Google Infrastructure Grant, and International Organization for Migration for support of this project.


References

Adekanmbi, D., Ayoola, F., & Idowu, A. (2014). Demographic time series modelling of total deaths in Nigeria. Southern African Journal of Demography, 15(1), 21–48.

AJMC. (2021, January 1). A timeline of COVID-19 developments in 2020. https://www.ajmc.com/view/a-timeline-of-covid19-developments-in-2020

Álvarez, R. (2020, September 25). Los bulos sobre la COVID-19 que defiende la plataforma ‘Médicos por la verdad’ y que ya hemos desmentido. [“The hoaxes about COVID-19 defended by the platform ‘Doctors for Truth’ that we have already debunked”]. Newtral. https://www.newtral.es/bulo-medicos-por-la-verdad-vacuna-coronavirus/20200925/

Armus, T. (2021, February 11). Among Latino immigrants, false vaccine claims are spreading as fast as the virus. The Washington Post. https://www.washingtonpost.com/local/latino-vaccine-misinformation-coronavirus/2021/02/11/40731dc6-64a3-11eb-8468-21bc48f07fe5_story.html?tid=ss_tw

Aro, J. (2016). The cyberspace war: Propaganda and trolling as warfare tools. European View, 15(1), 121–132. https://doi.org/10.1007/s12290-016-0395-5

Associated Press. (2020, July 19). Nicolás Maduro se disculpa por comentarios contra migrantes. [Nicolás Maduro apologizes for his comments about migrants.] AP News. https://apnews.com/article/8adbea7949bbb40cf8dcfa8019821bf5

Associated Press. (2021, January 27). Venezuela's Nicolás Maduro touts “miracle” cure for Covid, provides no evidence. NBC News. https://www.nbcnews.com/news/latino/venezuela-s-nicol-s-maduro-touts-miracle-cure-covid-provides-n1255835

Benton, A., Hancock, B., Coppersmith, G., Ayers, J., & Dredze, M. (2016, September 25). After Sandy Hook Elementary: A year in the gun control debate on Twitter. [Conference Presentation.] Bloomberg Data for Good Exchange Conference. New York City, NY, United States. https://data.bloomberglp.com/company/sites/2/2016/09/paper_15.pdf

Bode, L., Budak, C., Ladd, J. M., Newport, F., Pasek, J., Singh, L. O., Soroka, S. N., & Traugott, M. W. (2020). Words that matter: How the news and social media shaped the 2016 presidential campaign. Brookings Institution Press.

Bolgov, R., Chernov, I., Ivannikov, I., & Katsy, D. (2019). Battle in Twitter: Comparative analysis of online political discourse (Cases of Macron, Trump, Putin, and Medvedev). In A. Chugunov, Y. Misnikov, E. Roshchin, & D. Trutnev (Eds.), Communications in computer and information science: Vol. 1349. Electronic governance and open society: Challenges in Eurasia (pp. 374–383). Springer. https://doi.org/10.1007/978-3-030-13283-5_28

Bovet, A., & Makse, H. (2019). Influence of fake news in Twitter during the 2016 US presidential election. Nature Communications, 10(7), 1–14. https://doi.org/10.1038/s41467-018-07761-2

Box, G., & Jenkins, G. M. (1970). Time series analysis: Forecasting and control. Holden-Day.

Bozdağ, C. (2019). Bottom-up nationalism and discrimination on social media: An analysis of the citizenship debate about refugees in Turkey. European Journal of Cultural Studies, 23(5), 712–730. https://doi.org/10.1177/1367549419869354

Callaway, E. (2020, August 11). Russia’s fast-track coronavirus vaccine draws outrage over safety. Nature News. https://www.nature.com/articles/d41586-020-02386-2

Chaudhry, I. (2015). #Hashtagging hate: Using Twitter to track racism online. First Monday20(2). https://doi.org/10.5210/fm.v20i2.5450

Cho, V. (2001). Tourism forecasting and its relationship with leading economic indicators. Journal of Hospitality & Tourism Research, 25(4), 399–420. https://doi.org/10.1177/109634800102500404

Chung, R. C. P., Ip, W. H., & Chan, S. L. (2008). Impacts of the overheating economy on China’s manufacturing industry. The International Journal of Advanced Manufacturing Technology, 43, Article 1133. https://doi.org/10.1007/s00170-008-1792-y

Churchill, R., & Singh, L. (2021). textPrep: A text preprocessing toolkit for topic modeling on social media data. In Proceedings of the International Conference on Data Science, Technology, and Applications — DATA (pp. 60–70). SciTePress Science and Technology Publications. https://doi.org/10.5220/0010559000600070

Donato, K. M., Singh, L., Arab, A., Jacobs, E., & Post, D. (2021). Migration misinformation in Spanish-language Tweets during a pandemic. International Organization for Migration Research Series, No. 68. International Organization for Migration. https://publications.iom.int/books/mrs-no-68-migration-misinformation-spanish-language-tweets-during-pandemic

Dredze M., Broniatowski, D., & Hilyar, K. (2016). Zika vaccine misconceptions: A social media analysis. Vaccine, 34(30), 3441–3442. https://doi.org/10.1016/j.vaccine.2016.05.008

Duddy, P. (2015, March 30). Political crisis in Venezuela: Contingency Planning Memorandum Update. Council on Foreign Relations. https://www.cfr.org/report/political-crisis-venezuela

Ekman, M. (2019). Anti-immigration and racist discourse in social media. European Journal of Communication, 34(6), 606–618. https://doi.org/10.1177/0267323119886151

Ellsworth, B. (2021, March 26). Facebook freezes Venezuela president Maduro's page over COVID-19 misinformation. Reuters. https://reut.rs/3ffIOF2

Erdogan-Ozturkab, Y., & Isik-Guler, H. (2020). Discourses of exclusion on Twitter in the Turkish context: #ülkemdesuriyeliistemiyorum (#idontwantsyriansinmycountry). Discourse, Context & Media, 36, Article 100400. https://doi.org/10.1016/j.dcm.2020.100400

Fazio, L. K., Brashier, N. M., Payne, B. K., & Marsh, E. J. (2015). Knowledge does not protect against illusory truth. Journal of Experimental Psychology: General, 144(5), 993–1002. https://doi.org/10.1037/xge0000098

Fisher, M. (2020, April 13). Teorías de la conspiración del coronavirus: por qué prosperan y por qué son peligrosas. [Why Coronavirus Conspiracy Theories Flourish. And Why It Matters.] The New York Times. https://www.nytimes.com/es/2020/04/13/espanol/mundo/coronavirus-conspiracion-fake-news.html

Frenkel, S. (2021, March 10). La desinformación sobre las vacuna ataca a las comunidades hispanas y negras. The New York Times. https://www.nytimes.com/es/2021/03/10/espanol/vacunas-mitos.html

Gardel, L. (2020, September 23). ¿Quiénes son los “Médicos por la Verdad” y los “Epidemiólogos Argentinos”, grupos que difunden desinformaciones sobre el coronavirus? [Who are the "Doctors for Truth" and the "Argentine Epidemiologists", groups that spread misinformation about the coronavirus?] Chequeado. https://chequeado.com/el-explicador/quienes-son-los-medicos-por-la-verdad-y-los-epidemiologos-argentinos-los-2-grupos-que-difunden-desinformaciones-sobre-el-coronavirus/

Gerver, M. (2017). Misinformation as immigration control. Res Publica, 23, 495–511. https://doi.org/10.1007/s11158-016-9339-9

Gilmour, S., Degenhardt, L., Hall, W., & Day, C. (2006). Using intervention time series analyses to assess the effects of imperfectly identifiable natural events: A general method and example. BMC Medical Research Methodology, 6(16), Article 16. https://doi.org/10.1186/1471-2288-6-16

Guidry, J., Carlyle, K., Messner, M., & Jin, Y. (2015). On pins and needles: How vaccines are portrayed on Pinterest. Vaccine, 33(39), 5051–5056. https://doi.org/10.1016/j.vaccine.2015.08.064

Guo, L. (2020). China’s “fake news” problem: Exploring the spread of online rumors in the government-controlled news media. Digital Journalism, 8(8), 922–1010. https://doi.org/10.1080/21670811.2020.1766986

Haber, J. R., Singh, L., Budak, C., Pasek, J., Balan, M., Callahan, R., Churchill, R., Herren, B., & Kawintiranon, K. (2021). Research note: Lies and presidential debates: How political misinformation spread across media streams during the 2020 election. Harvard Kennedy School Misinformation Review 2(6), 1–38. https://doi.org/10.37016/mr-2020-84

Hasher, L., Goldstein, D., & Toppino, T. (1977). Frequency and the conference of referential validity. Journal of Verbal Learning and Verbal Behavior, 16(1), 107–112. https://doi.org/10.1016/S0022-5371(77)80012-1

Haynes, C., Merolla, J., & Karthick Ramakrishnan, S. (2016). Framing immigrants: News coverage, public opinion, and policy. Russell Sage Foundation.

Henkel, L. A., & Mattson, M. E. (2011). Reading is believing: The truth effect and source credibility. Consciousness and Cognition, 20(4), 1705–1721. https://doi.org/10.1016/j.concog.2011.08.018

Hernandez, S. (2018). #election#elección: Latino Twitter users and reactions to presidential political gaffes. In B. R. Warner, D. G. Bystrom, M. S. McKinney, & M. C. Banwart (Eds.), An unprecedented election: Media, communication, and the electorate in the 2016 campaign (pp. 369–382). Praeger.

Hernández-García, I., & Giménez-Júlvez, T. (2020). Characteristics of YouTube videos in Spanish on how to prevent COVID-19. International Journal of Environmental Research and Public Health, 17(13), Article 4671. https://doi.org/10.3390/ijerph17134671

Huang, Z., & Wang, R. (2019). Building a network to “tell China stories well”: Chinese diplomatic communication strategies on Twitter. International Journal of Communication, 13, 2984–3007. https://ijoc.org/index.php/ijoc/article/download/11221/2703

Hultkrantz, L., & Olsson, C. (1997). Chernobyl effects on domestic and inbound tourism in Sweden – A time series analysis. Environmental and Resource Economics, 9(2), 239–258. https://doi.org/10.1007/BF02441380

Human Rights Watch. (2020, October 13). Venezuela: Abusive treatment of returnees: Poor quarantine conditions may spread Covid-19. https://www.hrw.org/news/2020/10/13/venezuela-abusive-treatment-returnees

Hyndman, R., & Khandakar, Y. (2008). Automatic time series forecasting: The forecast package for R. Journal of Statistical Software, 27(3), 1–22. https://doi.org/10.18637/jss.v027.i03

International Crisis Group. (2020, December 14). Disorder on the border: Keeping the peace between Colombia and Venezuela. International Crises Group: Latin America & Caribbean: Report No. 84. https://www.crisisgroup.org/latin-america-caribbean/andes/colombiavenezuela/84-disorder-border-keeping-peace-between-colombia-and-venezuela

International Organization for Migration. (2021). Migration data in South America. Migration Data Portal. Retrieved February 6, 2022, from https://www.migrationdataportal.org/regional-data-overview/migration-data-south-america

Ip, W. H., Fung, R., & Keung, K. W. (1999). An investigation of stochastic analysis of flexible manufacturing systems simulation. International Journal of Advanced Manufacturing Technology, 15(4), 244–250. https://doi.org/10.1007/s001700050063

Isaac, L., & Griffin, L. (1989). Ahistoricism in time-series analyses of historical process: Critique, redirection, and illustrations from U.S. labor history. American Sociological Review, 54(6), 873–890. https://doi.org/10.2307/2095713

Jackson, B., & Jamieson, K. H. (2007). UnSpun: Finding facts in a world of disinformation. Random House.

Kreis, R. (2017). #refugeesnotwelcome: Anti-refugee discourse on Twitter. Discourse & Communication, 11(5), 498–514. https://doi.org/10.1177/1750481317714121

Kurmanaev, A., Herrera, I., & Urdaneta, S. (2020, April 19). Venezuela deploys security forces in coronavirus crackdown. The New York Times. https://www.nytimes.com/2020/08/19/world/americas/coronavirus-venezuela.html

Lai, S. L., & Lu, W.-L. (2005). Impact analysis of September 11 on air travel demand in the USA. Journal of Air Transport Management, 11(6), 455–458. https://doi.org/10.1016/j.jairtraman.2005.06.001

Lewandowsky, S., Cook, J., Ecker, U. K. H., Albarracín, D., Amazeen, M. A., Kendeou, P., Lombardi, D., Newman, E. J., Pennycook, G., Porter, E., Rand, D. G., Rapp, D. N., Reifler, J., Roozenbeek, J., Schmid, P., Seifert, C. M., Sinatra, G. M., Swire-Thompson, B., van der Linden, S., Vraga, E. K., Wood, T. J., & Zaragoza, M. S. (2020). The debunking handbook 2020. https://sks.to/db2020

Lodoño, E. & Simões, M. (2020, June 13). Brazil President Embraces Unproven ‘Cure’ as Pandemic Surges. The New York Times. https://www.nytimes.com/2020/06/13/world/americas/virus-brazil-bolsonaro-chloroquine.html

Lodoño, E., Casado, L., & Rasgon, A. (2021, March 6). As Covid deaths soar in Brazil, Bolsonaro hails an untested nasal spray. The New York Times. https://www.nytimes.com/2021/03/06/world/americas/brazil-covid-bolsonaro-nasal-spray.html

Martínez, L. (2020, August 20). Son falsas las afirmaciones de la inmunóloga Roxana Bruno, integrante de la agrupación “Epidemiólogos Argentinos.” [FALSE: The statements of the immunologist Roxana Bruno, a member of the group “Argentine Epidemiologists.”] Chequeado. https://chequeado.com/el-explicador/son-falsas-las-afirmaciones-de-la-inmunologa-roxana-bruno-integrante-de-la-agrupacion-epidemiologos-argentinos/

Maurer, M., & Reinemann, C. (2006). Learning versus knowing: Effects of misinformation in televised debates. Communication Research, 33(6), 489–506. https://doi.org/10.1177/0093650206293252

Méndez-Triviño, A. (2020, December 10). Forced back home by the pandemic, Venezuelan grandmother sees no choice but to flee once again. United Nations High Commissioner for Refugees. https://www.unhcr.org/en-us/news/stories/2020/12/5fd14a064/forced-home-pandemic-venezuelan-grandmother-sees-choice-flee-once.html

Oyeyemi, S. O., Gabarron, E., & Wynn, R. (2014). Ebola, Twitter, and misinformation: A dangerous combination? BMJ, 349, Article g6178. https://doi.org/10.1136/bmj.g6178

Ozduzen, O., Korkut, U., & Ozduzen, C. (2020). “Refugees are not welcome”: Digital racism, online place-making and the evolving categorization of Syrians in Turkey. New Media & Society, 23(11), 3349–3369. https://doi.org/10.1177/1461444820956341

Pardo, D. (2020, June 10). Coronavirus en Colombia y Venezuela | "Crucé todo un país a pie para nada": los venezolanos que intentan regresar a su país y cuyo gobierno ahora les restringe la entrada. [Coronavirus in Colombia and Venezuela: I crossed an entire country on foot for nothing": Venezuelans trying to return to their country and whose government now restricts their entry.] BBC Mundo. https://www.bbc.com/mundo/noticias-america-latina-52988879

Pranesh, R., Farokhenajd, M., Shekhar, A., & Vargas-Solar, G. (2021). CTMA: COVID-19 Misinformation Multilingual Analysis on Twitter. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing: Student Research Workshop (pp. 270–283). http://doi.org/10.18653/v1/2021.acl-srw.28

Pennycook, G., Cannon, T. D., & Rand, D. G. (2018). Prior exposure increases perceived accuracy of fake news. Journal of Experimental Psychology: General, 147(12), 1865–1880. https://doi.org/10.1037/xge0000465

Posadas-Durán, J.-P., Gómez-Adorno, H., Sidorov, G., & Moreno Escobar, J. J. (2019). Detection of fake news in a new corpus for the Spanish language. Journal of Intelligent and Fuzzy Systems, 36(5), 4869–4876. https://doi.org/10.3233/JIFS-179034

Ricard, J., & Medeiros, J. (2020). Using misinformation as a political weapon: COVID-19 and Bolsonaro in Brazil. Harvard Kennedy School Misinformation Review, 1(3). https://doi.org/10.37016/mr-2020-013

Ruokolainen, H., & Widén, G. (2020). Conceptualising misinformation in the context of asylum seekers. Information Processing & Management, 57(3), 102–127. https://doi.org/10.1016/j.ipm.2019.102127

Sanovich, S. (2017). Computational propaganda in Russia: The origins of digital misinformation. Computational Propaganda Research Project Working Paper No. 2017.3. Project on Computational Propaganda. https://demtech.oii.ox.ac.uk/wp-content/uploads/sites/89/2017/06/Comprop-Russia.pdf

Scheufele, D., & Krause, N. (2019). Science audiences, misinformation, and fake news. PNAS, 116(16), 7662–7669. https://doi.org/10.1073/pnas.1805871115

Shao, C., Ciampaglia, G. L., Varol, O., Yang, K.-C., Flammini, Y., & Menczer, F. (2018). The spread of low-credibility content by social bots. Nature Communications, 9(1), Article 4787. https://doi.org/10.1038/s41467-018-06930-7

Shao, C., Hui, P.-M., Wang, L., Jiang, X., Flammini, A., Menczer, F., & Ciampaglia, G. L. (2018). Anatomy of an online misinformation network. PLOS One, 13(4), Article e0196087. https://doi.org/10.1371/journal.pone.0196087

Sharma M., Yadav, K., Yadav, N., & Ferdinand, K. (2017). Zika virus pandemic-analysis of Facebook as a social media health information platform. American Journal of Infection Control, 45(3), 301–302. https://doi.org/10.1016/j.ajic.2016.08.022

Stanley, M. L., Yang, B. W., & Marsh, E. J. (2019). When the unlikely becomes likely: Qualifying language does not influence later truth judgments. Journal of Applied Research in Memory and Cognition, 8(1), 118–129. https://doi.org/10.1016/j.jarmac.2018.08.004

Wang, N., Varghese, B., & Donnelly, P. (2016). A machine learning analysis of Twitter sentiment to the Sandy Hook shootings. In 2016 IEEE 12th International Conference on e-Science (pp. 303–312). IEEE. https://doi.org/10.1109/eScience.2016.7870913

Wang, Y., McKee, M., Torbica, A., & Stuckler, D. (2019). Systematic literature review on the spread of health-related misinformation on social media. Social Science & Medicine, 240, Article 112552. https://doi.org/10.1016/j.socscimed.2019.112552 

World Health Organization. (2020, November 23). Consejos para la población acerca de los rumores sobre el nuevo coronavirus (2019-nCoV). https://www.who.int/es/emergencies/diseases/novel-coronavirus-2019/advice-for-public/myth-busters

World Health Organization. (2022, January 19). Coronavirus disease (COVID-19) advice for the public: Mythbusters. https://www.who.int/emergencies/diseases/novel-coronavirus-2019/advice-for-public/myth-busters

United Nations High Commissioner for Refugees. (n.d.). Venezuela Situation. Retrieved February 6, 2022, from https://www.unhcr.org/en-us/venezuela-emergency.html

U.S. Food and Drug Administration. (2022, February 1). Comirnaty Pfizer-BioNTech COVID-19 vaccine. https://www.fda.gov/emergency-preparedness-and-response/coronavirus-disease-2019-covid-19/pfizer-biontech-covid-19-vaccine

Van Duyn, E., & Collier, J. (2019). Priming and fake news: The effects of elite discourse on evaluations of news media. Mass Communication and Society, 22(1), 29–48. https://doi.org/10.1080/15205436.2018.1511807

Viscidi, L. (2016). Venezuela on the brink: How the state wrecked the oil sector—and how to save it. Foreign Affairs, 95(5), 133–140. https://www.foreignaffairs.com/articles/venezuela/venezuela-brink

Vosoughi, S., Roy, D., & Aral, S. (2018). The spread of true and false news online. Science, 359(6380), 1146–1151. https://doi.org/10.1126/science.aap9559

VTV Canal 8 [@VTVcanal8]. (2020, June 11). ¿Si son tan chéveres Brasil, Colombia, Chile, Ecuador y Perú, por qué los venezolanos no se quedaron allá? [If Brazil, Colombia, Chile, Ecuador and Peru are so cool, why didn't Venezuelans stay there?] [Video attached] [Tweet] Twitter. https://twitter.com/vtvcanal8/status/1271219754011852800


Appendix

Figure A1. Time-Series Plot of Newspaper Articles Discussing Maduro’s “Miracle Cure” Announcement


Table A1. Misinformation Themes, Topics, Myths, and Phrases

Below are detailed examples of the phrases used to measure the misinformation themes, topics, and myths related to COVID-19 and migration that we analyze in this article. For example, the vaccine myths theme captures incorrect ideas that include: the vaccine includes a microchip, its side effects are worse than COVID-19 itself, and it can alter someone’s DNA. The miracle cure theme includes mentions of miracle droplets, miracle cure, and Carvativir, for examaple, the chemical base of the supposed cure, and the border restrictions theme includes myths related to open borders, illegal crossings, and Táchira, a Venezuelan/Colombian border crossing station that the Venezuelan government often targets in its rhetoric about returning migrants 

Theme and Topic

Myth

Example topic phrases

COVID-19 Misinformation Conversation

Origins of COVID-19

Foreign virus*

Virus came from abroad; foreign virus

Origins of COVID-19

Chinese virus

Chinese virus; yellow virus; Wuhan virus; Chinese COVID

Origins of COVID-19

Migrant virus

Migrants bring COVID; Migrant virus; Migrants have COVID

Origins of COVID-19

Government hid COVID information

COVID government coverup; government COVID secrets

Origins of COVID-19

COVID was made in a lab*

COVID made in lab; COVID made in Wuhan lab

Origins of COVID-19

COVID was invented by humans

Humans made COVID; Humans invented COVID

Origins of COVID-19

COVID is a biological weapon*

COVID biological warfare; COVID biological weapon; bioweapon COVID

Origins of COVID-19

Pandemic is fake*

COVID was invented; Pandemic is fake; fake pandemic false pandemic

Origins of COVID-19

Pandemic was planned

‘Plandemic’; COVID was designed; COVID was planned

Transmission

COVID spread by cellular networks*

COVID cell phone spread; 5g spreads COVID

Transmission

COVID spread by mosquitoes*

COVID mosquitoes; mosquitoes spread COVID; mosquito bite COVID

Transmission

COVID is not spread in hot weather*

High temperatures prevent COVID; hot weather COVID spread

Transmission

COVID is not spread in cold weather*

Low temperatures prevent COVID; cold weather COVID spread

Transmission

COVID spreads HIV

COVID HIV; COVID spreads HIV

Transmission

COVID changes DNA

COVID DNA; COVID changes DNA

Transmission

Wearing mask can cause oxygen deficiency*

Mask oxygen deficiency; mask causes hypoxia

Prevention/treatment

Sun cures COVID

Sun cures COVID; Sun prevents COVID

Prevention/treatment

Snow cures COVID

Snow cures COVID; Snow prevents COVID

Prevention/treatment

Home remedies cure COVID*

COVID home remedies

Prevention/treatment

Bathing cures COVID

COVID showering; Taking a bath cures COVID

Prevention/treatment

Vitamins cure COVID*

COVID vitamins; COVID minerals

Prevention/treatment

Pepper cures COVID*

Pepper cures COVID

Prevention/treatment

Disinfectants cure COVID*

Disinfectants cure COVID

Prevention/treatment

Hot water cures COVID

Hot water cures COVID

Prevention/treatment

COVID can be treated hydroxychloroquine*

Hydroxychloroquine COVID

Prevention/treatment

COVID can be treated by Ivermectin*

COVID Ivermectin

Prevention/treatment

COVID can be treated by Dexamethasone*

Dexamethasone COVID

Prevention/treatment

COVID can be treated by anticoagulants*

COVID anticoagulants

Prevention/treatment

UV lights cure COVID*

UV lights cure COVID; UV lights COVID

Prevention/treatment

Bleach cures COVID*

COVID bleach; Bleach cures COVID

Prevention/treatment

Herbal tea cures COVID*

COVID herbal tea; Herbal COVID

Prevention/treatment

Methanol can prevent covid*

Methanol prevents COVID; Methanol COVID disinfect

Vaccine myths

Vaccine is ineffective*

Vaccine doesn't work; vaccine ineffective

Vaccine myths

Vaccine isn't proven to work*

Vaccine untested; Vaccine unproven; Vaccine not tested

Vaccine myths

Anti-vaccine*

anti-vaxx; antivaxx; antivaxxers

Vaccine myths

Vaccine has microchip*

Vaccine has microchip; vaccine microchip

Vaccine myths

Vaccine can track location*

Vaccine tracks location

Vaccine myths

Vaccine changes DNA*

Vaccine changes DNA; Vaccine DNA

Vaccine myths

Vaccine infects you with COVID*

Vaccine gives you COVID; Vaccine infects COVID

Vaccine myths

Vaccine availability*

Vaccine won't be available; vaccine unavailable

Vaccine myths

Vaccine approval*

Vaccine not approved

Vaccine myths

COVID cure*

Discovered COVID cure; They cured COVID

Vaccine myths

Vaccine cures COVID*

Vaccine COVID cure; Vaccine cures COVID

 COVID-19 Misinformation and the President of Venezuela

"Miracle cure"


Miracle droplets

Miracle droplets; Miracle solution; 100 percent effective; Oregano extract; Carvativir12

"Miracle cure"

Venezuelan scientists discover COVID cure

Venezuelan scientists cure COVID; Venezuelan scientists discover COVID cure

Migration Misinformation

Border restrictions

"Illegal" border crossings

Border illegal; illegal crossing; trochero13; crossing in the trocha14

Border restrictions

Military removal of "illegal" migrants

Ceofanb15 migrants; armed national forces illegal migrants

Border restrictions

Changes in border restrictions

Border reopening; border closure; border blocked; passage is closed

Venezuelans welcome back

Migrants will overwhelm Venezuela

Migrants overwhelm Venezuela; migrants are going to overwhelm our nation

Venezuelans welcome back

Migrant access to Colombian social services during COVID

migrants bono solidario16; Venezuelans ingreso solidario

Venezuelans welcome back

Venezuela is better than neighboring countries

other countries have xenophobia; Colombia coronahambre17

Migrant bioterrorists

Duque is infecting Venezuelans with COVID

Duque contaminates Venezuelans; Duque infects Venezuelans with COVID-19

Migrant bioterrorists

Migrants are Bioterrorists

Migrant bioterrorist; #denunciealbioterrorista18; migrants are a health threat

Migrant bioterrorists

Forced migrant quarantine

ceofanb covid migrants; ceofanb isolation centers covid; migrants isolation centers

Migrant bioterrorists

Misinformation about Venezuelan quarantine mandates

cuarentena radical19; 7+7; flexible week; radicalization of quarantine

*Final myths chosen based on how often they appeared in media and how dangerous they were from larger list of most common, salient, and dangerous Spanish-language myths.


Table A2a. Parameter Estimates of ARIMA Models for Vaccine Myths

Pre-event model
ARIMA(5,1,0)

Estimate (SE)

AR (1)

AR (2)

AR (3)

AR (4)

AR (5)

 

-0.510 (0.117)***

-0.494 (0.120)***

-0.328 (0.134)***

-0.489 (0.127)***

-0.387 (0.128)***

Post-event model
ARIMA(0,0,1)

 

MA (1)

0.397 (0.116)***

 

***p ≤ .001; **p ≤ .01; *p ≤ .05


Table A2b. Goodness of Fit Measures of ARIMA Models for Vaccine Myths

 

Pre-event model
ARIMA(5,1,0)

Post-event model
ARIMA(0,0,1)

AIC

BIC

RMSE

666.486

668.398

181.293

671.460

677.256

164.597


Table A3a. Parameter Estimates of ARIMA Models for Miracle Cure

Pre-event model
ARIMA(0,0,1)

Estimate (SE)

MA (1)

0.511 (0.122)*** 

Post-event model
ARIMA(4,1,0)

 

AR (1)

AR (2)

AR (3)

AR (4) 

-0.620 (0.125)***

 0.132 (0.156)***

-0.339 (0.148)***

-0.580 (0.168)***

***p ≤ .001; **p ≤ .01; *p ≤ .05


Table A3b. Goodness of Fit Measures of ARIMA Models for Miracle Cure

 

Pre-event model
ARIMA(0,0,1)

Post-event model
ARIMA(4,1,0)

AIC

BIC

RMSE

1973.892

1976.875

206.450

1950.171

1956.138

189.811


Table A4a. Parameter Estimates of ARIMA Models for Border Restrictions

Pre-event model
ARIMA(4,1,0)

Estimate (SE)

AR (1)

AR (2)

AR (3)

AR (4)

-0.441 (0.094)***

-0.605 (0.101)***

-0.214 (0.100)***

-0.360 (0.091)***

Post-event model
ARIMA(1,1,1)

 

AR (1)

MA (1)

 0.406 (0.085)***

-0.952 (0.032)*** 

***p ≤ .001; **p ≤ .01; *p ≤ .05


Table A4b. Goodness of Fit Measures of ARIMA Models for Border Restrictions

 

Pre-event model
ARIMA(4,1,0)

Post-event model
ARIMA(1,1,1)

AIC

BIC

RMSE

1963.845

1966.822

208.320

1948.262

1957.192

194.355

Note. ARIMA = autoregressive integrated moving average.


©2022 by Katharine M. Donato, Lisa Singh, Ali Arab, Elizabeth Jacobs, and Douglas Post. This article is licensed under a Creative Commons Attribution (CC BY 4.0) International license, except where otherwise indicated with respect to particular material included in the article.

Comments
0
comment

No comments here

Why not start the discussion?