Turning inward during crises: How COVID is changing our social ties

Ashwini Ashokkumar and James W. Pennebaker, The Pandemic Project and the University of Texas at Austin

Over the past few decades, Americans have experienced several disasters: the September 11th attacks, the explosion of the space shuttle Challenger, Hurricanes Katrina and Harvey, mass shootings at Sandy Hook and Las Vegas, wildfires, and many other painful losses.

After most of these disasters, we turned to family members, friends, and colleagues for comfort over the next days and weeks. Typically, the disasters occurred quickly, over a few hours or days, and our lives gradually returned to normal after a few weeks.

But the COVID outbreak is unlike any cultural upheaval in recent memory. The “enemy” is an invisible, contagious virus that spreads from person to person. Other people, even family and friends, have become a potential danger, and two months into the crisis, we remain isolated in our homes. Coming together with others not only isn’t a comfort — it’s a threat.

As we try to maintain an uneasy balance between infection and isolation, what’s happening to our social relationships and our sense of community? Are our social connections strengthened by these hard times or is the forced isolation weakening our ties to others? And, if we manage to connect with others, who do we feel closest to?

As part of the pandemic project, we conducted a series of surveys and Reddit analyses of thousands of people to understand how our social worlds changed during the first two months of the COVID crisis. More details on the method can be found here.

Our lab studies people’s language to understand their psychological experiences. In our previous post, we analyzed people’s natural language used in Reddit conversations to track their thoughts and emotions. Here, we examine people’s conversations to understand how connected they feel with various social groups.

If people are feeling more socially connected and thinking more about their social groups, they should make references to their connections with friends, family members, and others around them. Studies have found that after disasters such as 9/11, Hurricane Harvey, and the bonfire tragedy at Texas A&M, people talked more about their social relationships. What about during the COVID outbreak?

Using the language analysis program LIWC, we calculated how much each of hundreds of thousands of posts on Reddit communities used affiliation words such as we, us, our, together, and love, which indicate a focus on social relationships. As shown in the graph below, people began to use affiliation words more in the last week of February (for a detailed explanation of the timeline of COVID-19, read this post), indicating that they started to increasingly focus on their social relationships. Mid-march, around the time when shelter-in-place directives began, there was a sharp increase in feelings of social connection. Overall, COVID increased people’s feelings of social connection.

Figure 1. Affiliation words indicating a focus on social relationships (we, us, our, together) increased when the COVID crisis began, indicating that people are feeling more socially connected. February 29 was the day on which the first US death was announced. On March 11, the World Health Organization declared COVID to be an international pandemic.

The graph raises an interesting question: Who are people feeling connected with? Who are they talking about when they use words such as “we”? They might be talking about their parents, children, and other family members. Perhaps they are referring to friends they miss. They may also be talking about their city or even their country.

To answer this question, we turned to our survey data. In a series of large-scale surveys, over 15,000 people rated how much they felt connected to various groups relative to pre-COVID times. Respondents rated four groups — family, friends, city, and country — on a scale ranging from much less connected to much more connected. As seen in Figure 2, the only group that people reported feeling more connected to during COVID is family. In contrast, they felt less connected to their friends, city, and country.

Figure 2. Relative to pre-COVID times, survey respondents felt more connected with their family but less connected with their friends, city, and country.

We then went back to the Reddit data to test whether people’s conversations reflect the same pattern. Presumably, when people feel particularly connected to a group, they talk about it. Based on this reasoning, people should talk more than usual about their family but less than before about friends, city, or country. To test this, we analyzed the Reddit comments by counting the number of references people made to each of these groups.

We measured references to family by counting words such as father, mother, brother, and so on. The graph on the left of Figure 3 resembles the pattern in the use of affiliation words. Talking about family members increased around the time when the isolation period began in the second week of March. Using a similar method to track references to friends (the graph on the right), we found a drop in talking about friends beginning in the last week of February when COVID warning signs began followed by a second drop in March when isolation began.

Figure 3. People talked more about their family (left panel) and less about their friends (right panel) once the COVID crisis began. COVID has increased feelings of connections with family but weakened feelings of connections with friends.

Is this because most people are living with their family members during the current isolation period and spending more time with them? Our data suggests otherwise. Although almost 85% of the 804 survey respondents reported staying with their family members (including romantic partners) as opposed to only about 5% who were staying with their friends, the pattern shown in Figure 2 was the same for all participants regardless of who they lived with. In other words, even the people who lived alone or with their friends felt more connected to their families but not other groups relative to pre-COVID days.

Using similar analyses, we tracked how much the same people in the Reddit sample talked about their city and country. Overall, they talked much less about their city once the COVID outbreak began. The patterns for talking about country were similar but weaker. Nevertheless, there was no evidence that the COVID outbreak caused an increase in connections to the country.

COVID seems to be dampening people’s sense of community unlike other crises such as the September 11th attacks and Hurricane Harvey which brought people together. It’s easy to blame this on the broader political division in the United States. However, as NYT columnist David Brooks notes, after most epidemics, strife is rife. During the Cholera epidemic in Naples in the 1880s, people blamed each other and various groups for the spread of the virus. John Barry’s excellent book on the 1918 pandemic, The Great Influenza, points to tears in the social fabric during and after the outbreak of the Spanish flu.

Standing back now, over 100 years later, it is easier to understand why pandemics sowed discord in society. With death all around, little information about the virus, and no way to know who was a threat and who was not, people stayed home for months fearing the virus. Neighbors, people from other communities, and strangers from other cities and countries were particularly threatening. People suspected and blamed each other.

In periods of great uncertainty, humans (and other animals) naturally turn to family members for protection. Our data suggest that people are feeling closer to their families but withdrawing from others in the community. Unlike other collective upheavals such as hurricanes that bind society together, COVID is causing us to turn to our most fundamental relationships at the expense of more distal ones.

While strong familial ties are valuable, a dampened sense of community is not. Historical accounts of pandemics highlight the dangers of societal mistrust. How we fight our increasing urge to turn inward in the coming days will be pivotal in how COVID ends. Unlike in 1918, advances in medicine and technology equip us to make informed decisions. At the same time, we may avoid some of the darker social episodes from past pandemics through digital technologies which can help us to connect with others in our social worlds.

Notes for researchers: In all temporal graphs in this post, the y-axis represents 3-day moving averages. The survey findings in Figure 2 were significant at p < .001 and were replicated in a separate Prolific sample.

The Unfolding of the COVID Outbreak: The shifts in thinking and feeling

Ashwini Ashokkumar and James W. Pennebaker, The Pandemic Project and the University of Texas at Austin

Think of a major event in your life such as living through a hurricane, getting married, needing an operation. In the days and weeks before the hurricane, saying your vows, or going to the hospital, the ways you thought and behaved probably changed. You likely sought out relevant information. You talked with others who had had similar experiences. You probably felt nervous and, at the same time, ran simulations through your mind about what you would do during and after the event occurred. And, once the event happened, you likely discovered that your life changed in ways you never anticipated.

Similar shifts likely occurred with you when COVID-19 appeared on the horizon. By understanding how individuals approach significant life experiences, we can predict how entire communities and cultures do the same. The current project seeks to track how we all are thinking and talking about COVID as it unfolds over time. How has COVID shaped our emotions and thinking patterns, when did these effects begin and how are they currently evolving?

Let’s take a quick trip down memory lane. Way back on January 21, the first publicized case of COVID was reported in the U.S. More stories started emerging in February. And then on February 25th, world stock markets started to crash because financial experts saw a pandemic coming. The first death in Seattle was reported four days later. On March 11, the World Health Organization (WHO) declared COVID to be a global pandemic, and two days later, President Donald Trump declared a national emergency.

And then the world changed. Almost immediately, several cities and states announced lockdown orders, urging people to self-isolate. Universities, businesses, schools, and other organizations closed or began to operate virtually. Panic buying for toilet paper, red meat, and liquor were reported as everyone began sheltering in place.

Figure 1. Timeline of COVID in the United States

As we write this report at the end of April from our respective bunkers, many people are now looking out their windows with the expectations that they can venture out in the next few days or weeks or months.

What has happened to us over the last two months? Our research team has been tracking people through the analysis of over 750,000 Reddit comments made in eight city subreddits (or communities) posted between January 10, 2020, and April 10, 2020, including New York City, Seattle, Austin, Boston, Houston, Chicago, Los Angeles, and Portland. We’ve also analyzed surveys of over 10,000 people from snowball and Prolific.co samples primarily in the U.S. and Canada. For more information on our methodology, click here.

COVID IS COMING! COVID IS COMING!

When individuals get wind of a major threat to their lives, they begin trying to learn more about it. They search for information in news outlets, they talk with their friends, and they go to online communities to learn what others know. When the news stories began appearing in January about a new and deadly strain of coronavirus, a growing number of people began asking questions about it online.

The graph below tracks how much people in our Reddit samples used words such as COVID and coronavirus. The first surge occurred when the stock market dropped (around February 25) due to economic changes in China. COVID references first shot up in the subreddits of New York City (home of Wall Street and the New York Times), and two days later in Seattle, where the first COVID death was announced on February 29.

Although references to COVID also surged in the other city subreddits, the biggest spike for all cities was around March 11th when the WHO acknowledged that we were all heading into a worldwide pandemic. Interestingly, ever since Americans moved into the isolation phase, the rates of discussions about COVID in the city subreddits have remained high and relatively constant.

Figure 2. People talked about COVID more when warning signals emerged at the end of February and also when self-isolation orders were first announced. The y-axis depicts the daily average percentage of words about COVID in Reddit comments.

When rumors of a potential threat surface, people begin asking what the implications might be for them. How serious will the threat be? When will the virus come to their city? They begin imagining how they would deal with an unknown disease in their country or community. They become focused on the future.

Once the threat arrives, people shift from future focus to present focus, as you can see in Figure 3. As it became clearer that the COVID virus was going to affect their region, people began to live in the here-and-now. Which stores are open? Where can they get toilet paper and medicine?

Figure 3. Future-focused language increased first when the warning signals for COVID emerged. Present-focused language shortly followed as the COVID crisis increasingly became part of everyday life. Time-relevant language includes verbs or references to the future (e.g., will, tomorrow) or the present (e.g., present tense verbs such as is or am or words such as now). The y-axis depicts standardized values.

A major upheaval that people have not experienced can force people to try to make sense of the event. Typically, when people are trying to make sense of unexpected events, they often use words like realize, understand, meaning, and because. Using our text analysis program LIWC2015, we are able to count these cognitive processing words to get a sense of how people are actively working through issues associated with COVID.

As shown in Figure 4, people increasingly engaged in cognitive processing starting when the warning signals began. The first peak in cognitive processing coincides with the day when US stocks first dropped. Cognitive processing also increased in the days after people began to self-isolate.

Figure 4. Cognitive processing to make sense of the situation increased when warning signals emerged and also when self-isolation began.

Anxiety, Anger, and Sadness

Immediately after natural and manmade disasters, people feel more negative and less positive. Increased anxiety, depression, and post-traumatic stress symptoms are also common. We tracked people’s expressions of anxiety, anger, and sadness from before to the outbreak until mid-April.

As would be expected, the expression of anxiety-related words shot up immediately after the COVID warning signals began. A second spike followed self-isolation orders. Since mid-March, anxiety has been decreasing steadily, but anxiety levels even in the first week of April were much higher than prior to the warning stage.

Figure 5. Anxiety levels shot up at the beginning of both the warning and self-isolation periods

In separate surveys that were completed by several thousand people between March 19 and April 10, we asked people what they were most anxious about. As you can see in Figure 6, people were most nervous about family members becoming infected followed closely by the fear that they themselves might unknowingly infect others. In fact, anxiety levels about COVID were much higher than people’s concerns about the potential economic losses that COVID could possibly bring.

Figure 6. Ratings in response to a series of questions asking, “To what degree are you worried about…”. People responded along a scale ranging from 1 (not at all) to 5 (a great deal).

A common misconception is that natural and manmade disasters immediately make people hostile and angry, causing them to lash out at others. News media frequently interview police departments or family abuse shelters expecting large increases in violent crimes and abuse. Rarely do these numbers go up in the first weeks. In fact, reports indicate a drop in crime rates in the US and around the world.

Our findings are consistent with national trends. People’s use of anger-related words in the community subreddits dropped at the beginning of the warning stage and then again when people began to self-isolate (see Figure 7). Reports of strife and tension following previous disasters suggest that this pattern of lowered anger may not last long. Increases in gun sales suggest that people are anticipating increased conflict in the coming months. But then, gun sales with the election of Obama in 2008 surged as well but subsequent national violence statistics did not increase.

Figure 7. Anger levels dropped when the COVID outbreak began and still remain lower than baseline.

The anger patterns are a surprise for many people. In the social psychology literature, many researchers point to the powerful effects of attribution. If people are anxious and they don’t know why, they are likely to lash out at others. They know they feel bad and so they look around and pin their bad feelings on reasonable targets around them. A natural disaster or COVID is different. People feel anxious but they all know precisely why they are feeling bad. Their spouse might annoy them but there is simply too much evidence that their bad feelings are attributable to this scary virus.

However, in the weeks ahead, people will gradually stop thinking about COVID but are likely to remain anxious. The attribution literature would predict that they would start attributing their anxiety to others around them. In other words, once COVID starts slipping from the front of our minds, we will start getting angrier and lashing out at others at higher rates.

Anxiety jumping up is not surprising; anger dropping generally surprises people. How about sadness? If you look at Figure 8, the sadness results make perfect sense. We are living in a sad time: Joblessness is the highest in our lifetimes; the economy is miserable; some of our friends or relatives have gotten sick and some may have died from COVID. The pattern suggests that prolonged isolation might be gradually impacting people’s mental health and wellbeing.

Figure 8. Expressions of sadness have been gradually increasing since the self-isolation period began.

Where we are and where we’re going

We are now at the end of April 2020. We are at the end of the beginning. Most U.S. states and many countries in the northern hemisphere are just beginning to experiment with loosening their daily movement restrictions. We are all having conversations about the relative safety or danger of venturing out.

There is not a clear social-psychological roadmap. The few studies that have looked at other natural and man-made disasters hint that the next phase will be bumpy. We should expect elevations in hostility and back-biting. People will continue to feel anxious. There may well be significant diversions in the ways young and older people respond. People under the age of 30 will likely look at the statistics and think that they are relatively safe. Those who are older will continue to harbor healthy fears of contracting the disease and will be much more reticent to venture into public places.

One significant unknown is how COVID will continue to affect the economy. To the degree that consumption patterns remain stalled and job prospects are slow to return, the aftereffects of COVID may last months or years, perhaps long after the threat of the virus disappears.

Notes for researchers: In all temporal graphs in this post, the y-axis represents 3-day moving averages.

Locked indoors: What we do daily is related to mental health

Ashwini Ashokkumar and James W. Pennebaker, The Pandemic Project and the University of Texas at Austin

Some preliminary results are presented from an online survey of 5,626 American and Canadian respondents from the week of March 19-27, which was a week after the first shelter-in-place guidelines were imposed. Some of the noteworthy findings from the survey include:

• On average, respondents spent approximately 22 hours in their home or a home of a friend or family member, about 1.5 hours outside, and about 20-30 minutes in commercial establishments such as grocery stores.
• In the home, people reported spending 5.6 hours watching movies, reading, learning new skills, and other leisure activities; 4.3 hours cooking, eating, cleaning, and house and yard work; 3.9 hours on social media (Facebook, Twitter, email, texting); 3.7 hours on work-related tasks; and almost an hour exercising.
• Most social media time tended to be devoted to watching, learning, and communicating with others about COVID. In fact, approximately 17 percent of the sample was immersed in COVID-19 topics for at least 7 hours of their day. The more people are exposed to COVID-related information, the more depressed and lonely they felt, and the more alcohol they consumed.
• Overall, those who are coping the best with the COVID outbreak are people who are exercising at least 30 minutes a day, going outdoors, and talking with at least two friends or family members face to face or virtually.
• Those who are having the most difficult time coping tend to spend at least 5 hours or more watching television or movies, playing video games, or engaging in social media unrelated to COVID.

Introduction and overview

The Pandemic Project is seeking to track and understand people’s everyday social and psychological states over the course of the COVID-19 crisis. Social scientists have long been interested in disasters such as 9/11, the Columbine and other massacres, the Loma Prieta and other earthquakes, wildfires, assassinations, etc. In virtually all other upheavals, people’s natural reactions have been to get together, talk, hug, and cry. Many disaster victims have reported the power of being with others to share their fears, sadness, anger, and other emotions during and after the events. COVID-19 is unique in requiring people to stay locked in their own homes and to avoid human contact with others.

From a social science perspective, COVID-19 has the elements of a perfect storm. It is a deadly disease with no cure that is difficult to track. It is an invisible virus that is silently and mysteriously transferred from person to person without their knowledge, causing terror around the world. All of us, then, are potential victims and agents of the disease. The only prevention to help the loss of life is to avoid exposure to the virus by avoiding social contact. And social contact is one of the only proven remedies for reducing anxiety and depression. Locking down entire populations changes everything about daily life — including what people do with their time, the ways they interact with their friends and family, how they work and play. This also means that the economy is on the brink of a major downturn, even collapse, causing millions to lose jobs and/or incur huge financial losses.

During the first weeks of the outbreak and quarantine, only about 15 percent of people in our research knew of anyone who had been diagnosed with COVID and about 1 percent knew someone who had died. Nevertheless, COVID-19 was a common topic of most conversations, emails, or social media posts, reflecting the enormous social psychological effects of the crisis..

COVID-19 is an overwhelming social psychological phenomenon. How are people connecting with one another? How will they change in the weeks and months ahead? What coping methods are effective and ineffective? These are the questions that the Pandemic Project seeks to answer. In this first paper, we focus on how people are spending their time in their new lives starting about a week after the first shelter-in-place recommendations/requests/orders were announced. Our findings are based on results of an online sample of 5,626 people, mostly women (80.5%), with a college degree (79.8%), who were employed full time (53%), and who averaged 43 years of age. (For details of the study, click here)

How are people spending their time?

The most striking aspect of the COVID outbreak is that the vast majority of the world population stopped their daily routines and remained in their houses for weeks at a time. Most people outside the health care sector and other essential services either worked from home or were furloughed or laid off from their jobs. Restaurants, bars, and other gathering places were closed. In the few locations where groups of people were allowed (grocery stores, pharmacies, U.S. gun stores, and other essential services), social distancing norms were mandated. Given the new movement restrictions, much of the world’s population had to establish completely new daily routines.

In the survey, we asked participants how and where they had spent their previous 24 hours. Overall, the average person spent approximately 22 hours either at their own residence or the residence of a family member or friend. The remaining two hours were largely spent either outdoors or at grocery stores and supermarkets.

Daily activities

Participants were asked how much time they spent on a range of activities in the 24 hours before taking the survey. The five primary categories of interest included work, daily household routines, leisure activities, social media, and exercising.

As you can see in the first figure, people spent the most time on leisure activities such as reading, playing video games, watching movies or TV or learning new skills, averaging 5.6 hours per day. About 21% of the sample reported spending over 8 hours on leisure activities.

Daily routines such as eating, cooking, chores, and house maintenance was the second most frequent category, taking up 4.2 hours, indicating that people are using their time to organize their homes. This was not true of all people; about 19% of the sample spent less than 2 hours on daily routines.

Figure 1. Average time spent on various activities in the 24 hours before taking the survey

National time-use studies find that the average American usually spends about 2.1 hours each day on social media. In our sample, the number was almost twice that — about 3.9 hours per day. In fact, over 11% of the sample spent more than 8 hours on social media! Previous reports that over 50% of Americans use social media to get their news suggest that the current spike reflects an increase in consumption of news related to COVID.

Interestingly, people were spending only about 3.7 hours working remotely from home. Respondents who were not students and who were employed at the time of completing the survey worked for about 4.8 hours, marking a 33% drop from the average work hours for Americans pre-COVID.

Overall, our respondents spent about 45 minutes each day exercising. About 32% of the sample did not exercise at all. Those who did averaged one hour and 12 minutes exercising.

Depression, loneliness, and alcohol use

Large-scale upheavals, both natural and manmade, are followed by spikes in stress, anxiety, and depression that sometimes last several months after the event. The most powerful antidote to anxiety is social support. People usually cling to their loved ones during times of crisis. Being forced to stay apart presumably exacerbates the mental health problems resulting from the current crisis.

Respondents provided estimates of how long in their previous day they experienced feelings of depression and isolation, and the amount of time they consumed alcohol or other recreational substances. In our sample, about 30.2% reported feeling depressed for over an hour in the previous day, and about 14.8% experienced such feelings for over 3 hours (See Figure 2). The amount of time people felt alone was similar. About 20 percent of respondents reported drinking alcohol or taking recreational drugs for 1-3 hours or more in the previous day. The high numbers suggest that COVID may be having a meaningful impact on mental health.

Figure 2. Number of depressed hours in the previous day reported by survey respondents

Exposure to COVID-related information

Being exposed to traumatic images on television, social media, or in print may boost levels of fear, concerns about the future, and poorer coping. Some of the most compelling evidence for the dangers of repeated exposure to disturbing images has been reported by Roxane Silver and her colleagues at the University of California-Irvine. Over her career, she has studied people’s reactions to watching coverage of wildfires, the 9/11 attacks, beheadings, school shootings, and other horrifying events. Time after time, she has demonstrated that the more exposure people have to violent images, the more likely they will show signs of distress — often for weeks or months. Within days of orders for people to self-isolate in North America, Silver warned about the dangers of watching too much media coverage of the Coronavirus. She was right.

Our respondents estimated the time they spent reading about, watching, or communicating information about COVID. As you can see in the graph, over 45% of the sample reported spending at least 3 hours in their previous day dealing with COVID-related information, and about 17% spent 8 hours or more.

Figure 3. Number of hours in the previous day that survey respondents spent dealing with COVID-related information

What are the links between watching, reading, or talking about COVID and mental health?

Is it unhealthy to become immersed in watching and thinking about the COVID crisis? We asked people how worried they were about getting infected by the virus, dying from it, infecting others, and so on. Overall, people who spent more time dealing with COVID-related information reported higher levels of COVID-related worries.

More striking is that exposure to COVID-related information is associated with feelings of depression. Similar patterns are observed for feelings of loneliness and alcohol use.

Figure 4. People who spent more time dealing with COVID related information spent more hours feeling depressed.

Warning: we cannot be sure from our analyses whether dealing with COVID-related information causes depression. Our findings are correlational and do not provide information regarding the cause. It is equally possible that people who are more prone to depressive moods seek out COVID-related information. In fact, some compelling research on depression by Christopher Beevers and his colleagues who are in the Department of Psychology at the University of Texas at Austin find that people who are depressed often seek out negative and threatening information. Consequently, our findings may be showing that depressed people are driven to find COVID-related information more than others. However, other research suggests that being exposed to frightening and depressing information can make people more afraid and depressed. Can avoiding excessive news coverage of the COVID-19 help improve your mood and reduce symptoms of depression and loneliness? Our research can’t answer that question. However, you can figure out the answer yourself. If you are feeling anxious, afraid, or depressed and you are watching more than 3 hours of COVID news each day, try reducing your intake of COVID information and see if you feel better.

What other activities are related to well-being?

Of all the factors we have studied so far, exposure to COVID-related information is most strongly related to depression and loneliness, and, by extension the lowest levels of well-being. The analyses of our first wave of data collection point to other correlates of well-being worth considering. Recall that we asked people to estimate the number of hours they had spent on a variety of tasks in the 24 hours prior to completing the survey. We then compared how they used their time with a crude measure of well-being (the total number of hours people were free of feelings of depression and loneliness).

Spend a minute and think about the numbers in the table below. The three activities at the top are all associated with low well-being (and higher feelings of depression and loneliness). As a comparison, being immersed in COVID-related news and conversation is correlated about -.23 with well-being. Interestingly, the more time that people engaged in social media (e.g., Facebook, Twitter, Instagram), even when this did not involve dealing with COVID-related information, the lower was their well-being. Similarly, lower well-being is correlated with watching a great deal of television and playing video games (i.e., over 5 hours). These effects are not related to people’s age or their gender.

The more satisfying results tell us what kinds of behaviors are associated with greater well-being among our respondents. The three most robust dimensions are going outside, exercising, and talking with friends and family (both face to face and virtually by phone, email, or Zoom or other video).

*A negative correlation indicates a reverse relationship. So, for example, the people who spent more hours on social media tended to have the lowest scores on well-being. Conversely, the more hours that people spent outside, the higher their well-being.

In the coming weeks, we will be releasing reports on more analyses. How do individual differences in gender, age, ethnicity, education level, and political beliefs shape how people are coping with the COVID crisis? What are the psychological impacts on health workers? Is COVID bringing us together or tearing us apart? To receive updates, follow this blog.

Note to researchers: All the analyses presented in this report are preliminary. All effects presented are significant at p < .001 and were replicated in a Prolific sample of ~450 participants.