CHAPTER 1

Social Media, Sociality, and Survey Research

Joe Murphy, Craig A. Hill, and Elizabeth Dean,

RTI International

As survey researchers, we have long been concerned about the future of survey research. Beset by problems not of its own making, the survey research discipline faces unprecedented challenges because of declining data quality—stemming from, for example, falling response rates, inadequate sampling frames, and antiquated approaches and tools. Conducting a survey is, at its core, a social interaction between a researcher and a (potential) respondent—a conversation with a purpose. The current pace of technological change—and the way people communicate with one another—presages the upheaval of survey research as we know it.

Thus, survey researchers should be—must be—searching for improvements to the way their research is conducted. Survey researchers should be—must be—constantly scanning the landscape of technological and social change, looking for new methods and tools to employ. In this spirit, we have been, somewhat jealously, watching the explosion of social media.

In this book, we introduce the concept of the sociality hierarchy; that is, three levels of sociality inherent in the current (and future) state of person-to-person interactions using computing devices: (1) broadcast, (2) conversational, and (3) community. Survey researchers should recognize these levels when attempting to apply new social media tools to survey research. This book presents examples of how this application can be done and, perhaps more importantly, how survey researchers should think about applying these tools in the future as a complement to traditional survey research.

In this first chapter, we discuss the advent of social media in its many and varied forms, and we define it from the perspective of a survey researcher. We also show why survey researchers should be interested in, and vigilant about, social media. We introduce the concept of the sociality hierarchy for social media and show examples of each level or category.

Throughout the rest of the book, we explain (and show), at a more practical level, how survey researchers can use the data generated by social media at each level of the sociality hierarchy. Finally, we suggest particular vectors on which survey researchers might find themselves with regard to the use of social media data and tools as we move inexorably into the future.

WHAT IS SOCIAL MEDIA?

Millions of people have joined networks like Facebook and Twitter and have incorporated them into their daily lives, and at least partially as a result, communications between individuals and groups have changed in a fundamental way. Every day, billions of transactions occur over electronic systems, and within this stream are data on individuals' behaviors, attitudes, and opinions. Such data are of keen interest to those conducting survey research because they provide precisely the types of information we seek when conducting a survey.

The term social media has become ubiquitous. But what is social media? The term social suggests two-way interactions between people, which may be classified as one-to-one, one-to-many, or many-to-many. Media, or tools that store and deliver information, typically include materials that deliver text, images, or sound, i.e., mass media like books and magazines, television, film, radio, and personal media like mail and telephone.

Media can be further delineated into analog media, which contain data in a continuous signal or physical format, and digital media, which store information in a binary system of ones and zeros. The important distinction for our purposes is that digital media are typically stored or transmitted through computers or digital devices and can be disseminated via the Internet (aka the web).

The term social media most commonly refers to web-based technologies for communication and sharing over the Internet. There is no single agreed-upon definition of social media, but Scott and Jacka (2011, page 5) contend that social media is the set of web-based broadcast technologies that enable the democratization of content, giving people the ability to emerge from consumers of content to publishers. Social media involves the intensive use of electronic media for people in contact through online communities (Toral et al., 2009), but no agreed-upon definition exists for the concept of online community either (De Souza & Preece, 2004).

We propose a specific working definition of social media for the purposes of survey research: Social media is the collection of websites and web-based systems that allow for mass interaction, conversation, and sharing among members of a network. In this definition, social media has four defining characteristics: user-generated content, community, rapid distribution, and open, two-way dialogue (Health Research Institute, 2012).

Social media must be distinguished from other similar terms that may refer more to the technological or structural aspects of online systems. For instance, the web, built on the infrastructure of the Internet, contains myriad sites that are not part of the social media construct. These sites include, for instance, information resources without an interactive, discussion, or sharing component. Such resources typify web 1.0, which describes resources that allow users to view and consume information from the web but without necessarily sharing or interacting with the contents (Krishnamurthy & Cormode, 2008). Web 2.0, however, refers to sites with user-generated content such as videos, music, blog text, and photos (Anderson, 2007). Web 2.0 has allowed users to interact with the web and with other users and has permitted the consumer to become the creator of content (Asberg, 2009). Web 2.0 has been used as an umbrella term for web-enabled applications built around user-generated or user-manipulated content (Pew Internet & American Life Project, 2011). Ravenscroft (2009) considers web 2.0 to be the social and participative web that includes tools emphasizing social networking (e.g., Facebook, Bebo, and LinkedIn), media sharing (e.g., MySpace, YouTube, and Flickr) and virtual worlds (e.g., Second Life).

The distinction between social media and web 2.0 can be rather nebulous, but Figure 1.1 shows how social media fit into the evolution of popular media and technology. The figure presents analog, digital, web 1.0, and web 2.0 media types progressing over time from left to right. Analog media are listed with arrows connecting them to digital and web media that have evolved from or supplanted them. Diamonds represent media that have traditionally been employed for survey research including the following:

Mail, used for delivering paper-and-pencil surveys

Telephone, used for conducting survey interviews

Digital audio, such as WAV files, used for recording interviews for verification and data quality purposes (Biemer et al., 2000)

E-mail and homepages (now more commonly referred to as webpages) for web surveys, notifications, and panel maintenance

Market research online communities, which are private social networks and websites to conduct qualitative marketing research with a selected community of members.

Figure 1.1 Media types.

Media represented by hexagons in Figure 1.1 hold potential for survey research innovation moving forward and are, thus, the focus of this book. In this book, we present our research on these new modes and methods such as online games, multiply massive online games, and virtual worlds (Chapters 5, 6, 10, and 11); social networking platforms like Facebook and Twitter (Chapters 2, 3, 4, and 9); smartphone-enabled media like mobile apps for survey data collection (Chapter 7); and web 2.0-enabled research resources such as crowdsourcing (Chapter 8).

The dotted line in Figure 1.1 indicates which media we consider social media for the purposes of this book. It encompasses web 2.0 media but also includes some typically classified under web 1.0 that allow for user-generated content, rapid distribution, or open, two-way dialogue.

SOCIAL MEDIA ORIGINS

Consonant with our definition of social media as web-based, most social media originated—were born—in the 1990s. The explosion in popularity of the Internet during this time eventually led to its steady penetration into many aspects of life (Leberknight et al., 2012). Between 1993 and 1995, Internet service providers (ISPs) began offering access in most major U.S. cities (Scott & Jacka, 2011). America Online (AOL) became a popular service by mailing access discs directly to consumers, which allowed them to try (and subscribe to) the Internet. Once on the Internet, users could participate in social activities such as sharing their opinions broadly through individual homepages, participating in bulletin board discussions, and engaging in other activities. Widespread use of social media and web 2.0, though, did not proliferate until the 2000s. With the launch of MySpace in 2003, users had the means to control and share media easily on their own personal pages and comment on the contents of their contacts' pages. Scott and Jacka (2011, page 14) argue that if 2000–2004 was about building platforms and tools, 2005–2009 could be defined as the period of user adoption and the remarkable change in how users connect, converse, and build relationships. By 2006, YouTube and Twitter had launched, and Facebook cracked the 10-million user mark.

Since 2006, the function of the web has moved rapidly in the direction of user-driven technologies such as blogs, social networks, and video-sharing platforms (Smith, 2009). This user-generated content is becoming more prevalent across the web, with most sites now allowing users to publish opinions, share content, and connect with other users.

SOCIAL NETWORKING SITES AND PLATFORMS

Social media use has skyrocketed in the last several years. Participation in social networking sites ballooned from 5% of all adults in 2005 to 50% in 2011. Facebook, in particular, grew from 5 million users in 2005 to more than 900 million in 2012. Twitter continues its amazing growth with nearly 500,000 new accounts created per day (Health Research Institute, 2012).

Web 2.0 tools and technologies in particular have allowed collaboration and communication across boundaries (Schutte, 2009). At the core of web 2.0 are social media sites like Facebook, Twitter, LinkedIn, YouTube, and many others facilitating social sharing to an unprecedented extent. These sites can be defined as social network sites or social network services—web-based services that allow individuals to (1) construct a public or semi-public profile within a bounded system, (2) articulate a list of other users with whom they share a connection, and (3) view and traverse their list of connections and those made by others within the system (boyd & Ellison, 2007, page 211). On social network sites, individual users share knowledge, pose and solve problems, seek and offer advice, tell stories, and debate issues of interest (Toral et al., 2009). boyd and Ellison differentiate social network sites from networking as the latter term implies that relationships are being initiated between strangers. Social networking sites make networking possible, but it is not their primary function nor is it what makes them distinct from other forms of computer-mediated communication. The defining quality of social network sites, according to boyd and Ellison, page 211, is that they enable users to articulate and make visible their social networks.

Social networking sites host online communities of people who share interests and provide ways for users to interact, including e-mail and instant messaging services (Shin, 2010). Social network sites allow users to create profile pages with personal information, establish friends or contacts, and communicate with other users (boyd & Ellison, 2007). On popular sites like Facebook, users communicate via private messages or public comments posted on profile pages. Users also have the option to communicate via instant message, voice, or video. Early adopters of the sites tended to be young people, but an increasing proportion is older. And, indeed, extroverts may use social network sites more often (Sheldon, 2008) as these sites contain relatively public conversations between friends and contacts (Thelwall et al., 2010).

In the United States, 81% of adults go online and 67% of online adults visit social networking sites. Forty percent of those who own a cell phone use a social networking site on their phones (Pew Internet & American Life Project, 2013). Among teens aged 12 to 17, 93% are online and 74% have a profile on a social networking site; social networking is the online activity on which they spend the most time daily (Pew Internet & American Life Project, 2011; Kaiser Family Foundation, 2010). Popular social networking sites and platforms take many different forms, as described in the next section.

Blogs

Blogs (or web logs) are periodically updated sites managed by an individual or group to provide information or opinion on a range of topics. Entries, or posts, are organized so that the most recent entry appears at the top. Many blogs allow and invite discussion from readers in the form of comments directly on the site. In popular blogs, the discussion via comments can contain more varied information and opinion than the original blog post itself. Blogs vary in length, frequency of post, topic, and formality. Because of the advent of free and easy-to-use web publishing tools (e.g., Blogger, WordPress, and Tumblr), blogs are available and used as an outlet for those with a wide range of resources and proficiency with computing. More than 150 million public blogs exist on the web (NielsenWire, 2012). Compared with the general population, bloggers are more likely to be female, in the 18- to 34-year-old age group, well-educated, and active across social media.

Blogs are often interconnected, and writers may read and link to other blogs, providing links to others in their entries. Often referred to as the blogosphere, interconnected blogs are social media in their own right, with social communities and cultures of their own (Mazur, 2010).

Twitter

Twitter has been defined as a microblogging service. Launched in 2006, Twitter users or Tweeters publish short messages, or Tweets, up to 140 characters in length. Tweets are publicly visible through the website or third-party applications in real time with millions of Tweets posted across the world per hour. Tweets contain status update information, sharing of links and information, direct messages to other Tweeters, and opinions on almost any topic imaginable (Tumasjan et al., 2010).

With the rise of text-based social media like Twitter and millions of people broadcasting thoughts and opinions on a variety of topics, researchers have taken notice. Twitter, in particular, is convenient for research because of the volume of publicly available messages and the process of obtaining them is relatively simple (O'Connor et al., 2010).

Survey research applications for Twitter are described in Chapters 2, 3, and 9.

Facebook

Launched in 2004, Facebook is a website and social networking service with almost 1 billion active users as of 2012, making it one of the most popular sites on the Internet. Enrollment is free, and users can create a profile and share information such as hometown, current city, education, employment, interests, and favorites. Users can also post photos, videos, notes, and status updates to share with their Facebook friends. Through Facebook, a user can make new contacts and follow, or like different groups, organizations, or products. A survey of 1,487 adults in the United States found that 43% of U.S. adults use Facebook; use is highest among young adults, college graduates, and those with an income greater than $90,000 (Morales, 2011).

Facebook users and their friends comprise virtual communities linked by shared interests or opinions. Researchers can sample populations of interest working through existing groups or creating new ones for the purposes of their analysis (Bhutta, 2012). As the number of Facebook users worldwide continues to grow, survey researchers are interested in how we might leverage social media to connect individuals and share information with one another (Lai & Skvoretz, 2011). Potential survey applications using Facebook are described in more detail in Chapter 4.

LinkedIn

Often considered the business world's version of Facebook, LinkedIn is a social networking service for professionals. With more than 150 million users in over 200 countries and territories, LinkedIn provides a means for individuals in various professional fields to interact and discuss issues, build their personal networks, and promote their expertise when seeking new opportunities or employment. Groups serve as hubs of interaction with sharing of announcements, links, and opinions. Over 47 million unique users visit LinkedIn monthly with more than 21 million from the United States.

Second Life

Second Life is an online three-dimensional world in which users (called residents in Second Life) design personal avatars and interact with other avatars and their surrounding environment. Communication in Second Life can be through instant messages or voice chat. As opposed to other social media sites like Facebook and Twitter, which typically augment real-life personas and relationships, residents in Second Life represent themselves in ways that depart from real-life appearances and personalities.

Second Life residents come from more than 100 countries. Second Life use is measured in user-hours—and 481 million user-hours were logged in 2009 (Linden, 2011). The most active users (as of 2008) were 25–44 years old (64% of hours logged) and male (59% of hours logged). Second Life provides a context-rich environment for conducting cognitive interviews and other survey pretesting activities (Dean et al., 2009; Murphy et al., 2010). The system allows the researcher to target and recruit specific types of residents through classified-type advertisements, online bulletin boards, and word-of-mouth in the virtual world, which can be more efficient and cost-effective when compared with traditional newspaper ads or flyers that are typically used to recruit in-person cognitive interview subjects (Dean et al., 2011). Text-based chat and voice chat can be collected for analysis, offering full transcripts of cognitive interviews (see Chapter 5). The only elements of in-person cognitive interviews missing from the paradata are facial and physical expressions, although Second Life residents can manipulate these for their avatars to a certain extent. Second Life is also useful in accessing hard-to-reach populations such as those with chronic illnesses (see Chapter 10) and for conducting methodological research (see Chapter 6).

Other Social Networking Platforms and Functionalities

A multitude of other social media platforms and functionalities exist, and the list grows continually. YouTube (owned by Google), for example, is a website for sharing videos. Users can upload original content (e.g., video blogs and performances) or media clips and discuss content on the site using comments. Flickr, Snapchat, and Instagram are popular sites with a similar design but primarily for posting and sharing photographs.

Another social networking site, Foursquare, focuses on physical locations. Users check in with their global positioning system (GPS)-enabled smartphones at various locations and earn awards in the form of points and badges.

For the purposes of this volume, we consider text messaging and smartphone applications (i.e., mobile apps) (although not technically web-based) to fall under the umbrella of social media. Text messaging can be used to interact with study respondents to capitalize on the cultural behaviors of the younger population. A text message is a short communication between mobile phones on a bandwidth lower than that of a phone call, and messages are usually limited to 160 characters (Cole-Lewis & Kershaw, 2010). Mobile apps are standalone pieces of software that a user can download to his or her device that allow for communication and interaction in many ways similar to web-based social media platforms. Chapter 7 discusses mobile apps in more detail.

WHY SHOULD SURVEY RESEARCHERS BE INTERESTED IN SOCIAL MEDIA?

The Current State of Survey Research

Survey research, as a discipline, has been under attack for the better part of several years now because of several coincidental trends that imperil the ability of a survey to provide data that are fully fit for use—accurate, timely, and accessible (Biemer & Lyberg, 2003, p. 13). One could argue that, in this day and age of advancing technology, survey data can be made timelier and more accessible quite easily—but if those data are not accurate, they are not of high quality and are not fit for use.

These coincidental developments and trends have resulted in an uneasy state of affairs for survey researchers. We now confront declining data quality as a result of, among other issues, falling response rates and inadequate sampling frames.

Falling Response Rates

Virtually everyone in the survey research industry knows that, regardless of mode, survey response rates (the number of completed interviews divided by the eligible sample population) are in decline. Scholars in the survey research field have been decrying the decline since at least the 1990s (de Heer, 1999; Steeh et al., 2001; Tortora, 2004; Curtin et al., 2005). That outcry continues to this day: In his presidential address delivered to members of the American Association for Public Opinion Research (AAPOR) in May 2012, Scott Keeter called the problem of declining response rates formidable and noted that they are well below the norms we were trained to expect. Furthermore, he cited response rates in the single digits in the public opinion (polling) sector of the industry even when conservatively calculated (Keeter, 2012, page 601).

Why are falling response rates a problem or threat to traditional survey research? Data quality has long been a cornerstone of social science research: Researchers are constantly investigating methods to maximize data quality and minimize survey error, which has many components and causes. The failure to obtain data from all sample members, referred to as unit nonresponse, can lead to bias in a researcher's estimates and, thus, flawed conclusions resulting from analysis of survey data if systematic differences exist in the key survey outcomes between responders and nonresponders (Peytchev, 2012). If, for example, the underlying cause for increased unit nonresponse is the topic of the survey (e.g., many sampled respondents refuse to participate in a health survey because they view their own health as a sensitive topic), the level of unit nonresponse (the response rate) will almost certainly be correlated with key survey estimates (Groves et al., 2006).

Although researchers and analysts can assess nonresponse bias and its impact in multiple ways, the response rate is the most well-known and most frequently used indicator of error stemming from unit nonresponse. It has been, and remains, an overall indicator of data quality—interpreted by many as a shorthand way for labeling whether a particular survey is good, scientific, valid—or not. (Carley-Baxter et al., 2006, page 2).

A combination of factors has likely led to survey researchers' increasing inability to achieve adequate response rates, including increased suspicion regarding requests for data from both government and corporations, increased reluctance to share personal data with unknown persons or entities, and tightened controlled-access to housing units, among others.

Some of the mistrust by potential respondents is no doubt because of burgeoning bids for their attention (and answers/data) from every conceivable angle: The amount of junk mail (both print and electronic) is rising. Telemarketers discovered, early on, the power and ease of calling people at home to ply their wares—so much so, that Americans were inundated with calls that were indiscernible at first from survey requests (Tourangeau, 2004). Some people might have sought to escape the onslaught by declining to answer the door or phone, and instead seeking refuge in their computer, but then spam e-mail and phishing attacks became more prevalent than ever (Kim et al., 2010; Tynan, 2002). The threat of computer viruses from unknown sources, and news stories about identity theft and stolen laptops containing confidential information likely led people to become more protective of their personal information, including refusing to cooperate with a survey solicitation—not bothering to make the distinction between a legitimate survey and any other entreaty for personal information.

People now have all sorts of ways by which they can avoid unwanted solicitations, including requests to participate in surveys. Mobile phones come equipped with caller ID. If people do not recognize the number, they can easily reject the incoming call. Such technology has certainly had an impact on surveys, contributing to declines in response rates and increasing the costs of conducting telephone surveys (Kempf & Remington, 2007; O'Connell, 2010).

Challenges with nonresponse have not been limited to telephone surveys. In-person (or face-to-face or field) surveys have been the gold standard for survey research since at least the 1940s, and most of the massive data collection efforts required by the U.S. federal statistical system are conducted this way (Yeager et al., 2011). But even gold standard surveys are experiencing downward-trending response rates. One reason is the increase of controlled-access housing developments, such as gated communities and buzzer systems in multiunit housing, making it much more difficult for even the most persistent interviewer to contact potential respondents (Keesling, 2008). In recent years, government-funded studies have documented an increase in the amount of nonresponse attributed to controlled access (Cunningham et al., 2005; Best & Radcliff, 2005). Because individuals have been able to restrict access in these ways, reaching them and collecting quality survey data has become more difficult and more costly (Curtin et al., 2005).

Frame Coverage Errors

A second major potential source of error in survey research—and, thus, another nail in the coffin being built by its detractors—is the ever-increasing number of cell-phone-only households and individuals. Surveys conducted by landline telephone, using a random-digit-dial (RDD) sampling frame, run the risk of missing entire segments of the population of interest if the traditional landline-based telephone sampling methods are not combined with a cell-phone frame (Blumberg & Luke, 2009). The latest data from the U.S. federal government show that more than a third (36%) of American adults now has a cell phone but no landline (see http://www.cdc.gov/nchs/data/nhis/earlyrelease/wireless201212.pdf or Blumberg et al., 2012). Some subgroups are more likely to fall in the cell-phone-only category: adults aged 25–29 (58.1%) and renters (52.5%), for example. These holes in sampling frame coverage have led some researchers to suggest that we abandon telephone sampling and telephone surveys entirely (Holbrook et al., 2003) or that the entire industry needs to be remade (O'Connell, 2010).

Other scholars are trying to develop alternative schemes for sampling, acknowledging that something new must be done (see Brick et al., 2011; Link et al., 2008). In this book, we suggest that many social science researchers are now moving beyond alternative sampling schemes and are instead exploring entirely new models of social science data acquisition and analysis. At least some of this search for new ways to find and interpret data is driven by the ready accessibility of data itself: The era of Big Data has begun.

The Coming Age of Ubiquity

In Alvin Toffler's The Third Wave (1980), he writes that there have been three waves of civilization, beginning with the agricultural wave, followed by the industrialization wave, and now supplanted by the information wave. Toffler further argues that the advent of each wave brings with it uncertainty, upheaval, and change. Those willing to surf along on top or ahead of the new wave—that is, embrace the change—can not only acquire power and wealth, but also fashion fundamental alterations of civilization.

The ubiquity of computers and technology has been the stuff of science fiction for decades. This ubiquity seems tantalizingly close now, but they are still just out of reach. The day is coming when your refrigerator knows that it is out of milk and signals the local grocery to add a gallon of 2% milk to your next order. We will not, however, have achieved this state of ubiquity until we no longer take notice of the role that computers and technology play in our everyday lives; that is, implicit in the notion of ubiquity is the idea that computers will perform these functions as a matter of course and we will have accepted that.

Variety. Transactions and interactions that are heavily computer dependent already occur on a daily basis. Making a phone call, sending a text message (SMS), making a bank deposit, swiping a loyalty card at Starbucks or Walmart, paying for a meal with a credit card, passing in front of a CCTV camera in London, posting a video on YouTube, querying Google, Tweeting on Twitter, and a hundred other seemingly innocuous activities that we all perform every day all involve a vast network of computers humming away behind the scenes, generating vast amounts of data. All of these bytes of information posted or transacted each day, taken together, equal Big Data.

Volume. As computers and computer-based technology near ubiquity, the amount of transactional and interactional data created grows exponentially. In 2012, every day 2.5 quintillion bytes of data (1 followed by 18 zeros) are created, with 90% of the world's data created in the last two years alone, according to one estimate (Conner, 2012, paragraph 3). In February 2011, Facebook users uploaded, on average, 83,000,000 photos every day. Walmart generates 1 million customer transactions every hour. Every minute (http://mashable.com/2012/06/22/data-created-every-minute/), for example, there are:

47,000 app downloads from the Apple App Store

3,600 pictures uploaded to Instagram

27,778 new posts on Tumblr

204,166,667 e-mails sent

Velocity. Computers operate at an ever-increasing pace, contributing to the generation of more and more data. Any particular set of variables of interest may now be analyzed on a daily basis, or even hourly or by the second, if preferred. Faster data processing and faster data transmission rates make for shorter and shorter feedback loops, making data and data analysis timelier than ever before.

Validity. However, we know that social media has produced much superfluous material as well. People create Facebook pages for their cats, sensors capture false positives, and data elements go missing. One of the biggest roles that social science researchers can play in this arena is to attempt to make sense of it all: Revolutions in science have often been preceded by revolutions in measurement, and the explosion of data and data accessibility demands new ways of measuring human behavior (Cukier, 2010).

There is a distinction to be made between the organic (Groves, 2011)