The events of 2011 transformed the politics of the Arab world. In just under 12 months, dictators fell in three countries: Tunisia, Egypt, and Libya. In Bahrain, Yemen, and Syria, protests and violence continued throughout 2012 to shake what were once assumed to be the solid foundations of long-standing autocratic regimes. Although it is too soon to assess the outcome of these events, one immediate consequence is clear: the empowerment of public opinion in Arab politics. For the first time in more than 50 years, the voices of average Tunisians, Egyptians, and Libyans can directly influence political outcomes. This shift may yet prove to be temporary, but its importance in driving current events cannot be discounted.
These political shifts and the subsequent empowerment of public opinion have outpaced regional polling efforts, injecting further uncertainty into an already unpredictable period of change. Discrepancies between electoral results and polling outcomes prior to both the parliamentary and presidential elections in Egypt are illustrative of this challenging dynamic (Abu Dhabi Gallup 2011; al-Ahram Online 2012; Telhami Reference Telhami2011). In 2011 and 2012, politicians campaigned and Egyptians voted without reliable measures of public attitudes toward the various parties and individual candidates. This situation helps to explain, at least in part, the shock in both the United States and Egypt at the success of the Salafi Hezb al-Nour party in the first round of parliamentary elections in 2011 and Ahmed Shafiq in the presidential elections of 2012. In the two years since the Egyptian revolution, several local organizations have attempted to initiate regular polling efforts that inform the public and the candidates about the opinions of Egyptian society (ʿAbd al-Gawad et al. Reference ʿAbd al-Gawad, Qamha and ʿAsila2012; al-Ahram 2011; DEDI 2011; Thawra Stats 2011), but many challenges clearly will have to be overcome before reliable polling becomes a routine part of Egyptian politics.Footnote 1
The increasing use of the Internet to search for information about candidates, parties, and events, however, suggests an intriguing alternative: that political scientists might use online search data to enhance conventional studies of public opinion in places like Egypt where traditional sources of such data are absent or unreliable. In the following sections, I review how search data has been used in the past, examine the advantages and disadvantages of using search data to gauge public opinion during periods of rapid transition, and illustrate these dynamics by examining how search data and the concept of “attentiveness” help to shed light on the battle for framing and narrative control in post-revolutionary Egypt.
PREVIOUS APPLICATIONS OF SEARCH DATA
Previous work has established the utility of Internet search data in highly connected societies. Researchers in economics and public health have been using Google's “database of intentions”Footnote 2 for several years to further our understanding of phenomena as varied as the spread of swine flu and other illnesses (Carneiro and Mylonakis Reference Carneiro and Mylonakis2009; Ginsberg et al. Reference Ginsberg, Mohebbi, Patel, Brammer, Smolinski and Brilliant2009), the retail prices of cars and homes (Choi and Varian Reference Choi and Varian2009), and revenues for video games and movies (Goel et al. Reference Goel, Hofman, Lahaie, Pennock and Watts2010). These studies are based on a simple observation: frequently a high correlation exists between online searches and offline outcomes. That discrepancies may exist between online searches and some events offline, or that only a fraction of the billions of searches in any given day pertain to a term of interest, is not a major challenge for these studies. To date, these discrepancies have been a limited concern because most researchers who utilize search data are concerned with developments in North America, East Asia, and Europe—regions characterized by high rates of Internet penetration. Higher levels of connectivity are crucial because they narrow potential demographic discrepancies between Internet users and the rest of society.Footnote 3
Furthermore, a variety of researchers, particularly in the fields of economics, sociology, and public health, have demonstrated that frequency of search and changes in search volume can be validated by more conventional measures of attitudes and behavior (Newig Reference Newig2004; Ripberger Reference Ripberger2011). For example, scholars who have looked at search data as a proxy tool for measuring public opinion have found high degrees of correlation with “Most Important Problem” (MIP) surveys (Mellon Reference Mellon2011; Scharkow and Vogelgesang Reference Scharkow and Vogelgesang2011). Not surprisingly, researchers also found search data to be an especially useful tool for capturing behavior or attitudes that respondents might be hesitant to express to an interviewer with a clipboard (Stephens-Davidowitz Reference Stephens-Davidowitz2012). Whereas the use of Internet data is far more limited within political science, the contributions to this symposium, as well as by other prominent researchers, suggest the significant potential of Internet data even in authoritarian and restrictive research environments (Chestnut-Greitens Reference Greitens2013, this issue; Hassanpour Reference Hassanpour2011; King, Pan, and Roberts Reference King, Pan and Roberts2012).
SEARCH DATA: ADVANTAGES
What use does search data have in a country like Egypt, where access to the Internet is far more limited than in East Asia, Europe, or North America?Footnote 4 At least three significant facets of search data make it a promising analytical tool for highly dynamic political environments where traditional survey methods may falter—even if Internet penetration rates in the developing world are much lower than the countries where such techniques have typically been applied before.Footnote 5
Format, Accessibility, and Cost
Internet search data from Google is free and easy to access. Its measurements are precise and uniform. Except for anomalous circumstances like the Internet blackout in Egypt January 27 to February 2, 2011 (Cowie Reference Cowie2011), researchers will always have access to search data. Reliable access to data is particularly important in a developing country like Egypt where, as one entrepreneur acknowledged, “you are conducting research in a country with no baselines.”Footnote 6 “Official” statistics in Egypt are often dubious, and even reliable data, when it is available, is frequently in a format that is unwieldy or expensive to digitize. Under these circumstances, a well-managed, publically available dataset is a useful asset and far less costly than conventional survey instruments (Granka Reference Granka2010; Scheitle Reference Scheitle2011).Footnote 7
Speed of Data Collection
Another advantage is that search data is up to the minute. Google's AdWords tool provides a 30-day rolling average of any search term. Google Trends, which provides normalized variations in search trends over time, updates its results, with only a two-day lag and can be filtered by date and location. This is a significant advantage when fast-paced events can often confound efforts to gauge attitudes toward political figures, parties, and developments.Footnote 8
Volume
Search data's most important advantage lies in its sheer volume. As of January 2013 approximately 32.8 million Google searches are made in Egypt on any given day.Footnote 9 Google is ubiquitous in Egypt, accounting for between 97% and 98% of the search market (Global Stats 2012). Google's dominant market share makes Google-derived data an extremely useful tool for identifying trends in search for Egypt's nearly 25 million estimated Internet users who themselves represent an important political constituency as well as a growing segment of Egyptian society (MCIT 2012).Footnote 10
SEARCH DATA: CHALLENGES
Several significant methodological limitations are inherent in the use of search data.
Probability Sampling
The most important advantage of surveys over search data concerns representativeness. An experienced research team will construct a random sample for their survey so the findings of the poll can be used to generalize about a larger population.Footnote 11 Because no observable “individual” exists in Google's summarized search data, a probability sample cannot be constructed. This makes statistically valid generalizations about the Egyptian population using publically available search data impossible.Footnote 12
Demographic Discrepancies
In developing countries like Egypt, where Internet penetration levels are relatively low, a second challenge emerges because of the potential for significant demographic discrepancies between Internet users and the rest of the population. For example, a recent survey suggests that Internet users in Egypt are disproportionately young and male, with the 15–34 age group strongly represented online (Insights MENA 2011). These users are also presumably wealthier, more focused in urban areas, and more likely to be literate than the average Egyptian, but no publically available data on the socioeconomic characteristics of Internet users in Egypt definitively confirms these assumptions. Finally, whereas surveys typically capture key demographic and political data (race, age, religious identity, political affiliation, etc.), search data, which simply tracks frequency of search terms over time, cannot offer such insights. This makes arguments about causality or controlling for specific demographic features impossible with publically available search data alone.
Nuance
A well-written questionnaire avoids misleading questions, ambiguous language, or focus on overly detailed issues of little relevance to the general public, also known as the “problem of nonattitudes” (Asher Reference Asher2007). By circumventing these pitfalls of survey design, questionnaires can provide a high degree of resolution on participants' attitudes. For example, surveys typically feature a variety of questions that allow responses to be placed across a spectrum. This creates a far more nuanced set of results even concerning complex topics like the implementation of “hadud” punishments or “views of democracy”—two examples from a Pew Global Attitudes Project survey in Egypt (Pew 2010).
Search data, unlike polling or a survey, cannot provide direct insights into the opinions behind the query. Someone could be searching for Mohammed Elbaradei because they support the former International Atomic Energy Agency (IAEA) head or because they loathe him. This example demonstrates a final challenge of aggregated search data—while spikes in search data are almost always a reflection of actual events, trend lines do not indicate whether this attention is positive or negative.Footnote 13 In other words, rarely are the motives behind any individual search apparent, although more sophisticated technical analysis might help tackle this challenge.
CASE STUDY: “EGYPT IS NOT TAHRIR”
“Egypt is not Tahrir or Muhammed Mahmoud Street …”
Major General Mukhtar El MallahFootnote 14The Utility of “Attentiveness:”Footnote 15 Although search data cannot be used as a proxy measure for intentions, it is a great measure of attention (Ripberger Reference Ripberger2011).Footnote 16 Three components of the “attentiveness” concept, as captured by search data, are of immediate interest to political scientists. First, unlike a poll that captures attitudes, search data measures behavior. Second, because searching for something online requires a time commitment, it is also presumably a reflection of genuine interest. Finally, because the person entering the search term is not under observation by a researcher, there are few constraints on what they search (see Stephens-Davidowitz Reference Stephens-Davidowitz2012). These three factors are what give the concept of “attentiveness,” first developed by Ripberger with regards to search data, its analytical punch. When we use search data, we are examining the aggregate actions of individuals, each at a distinct time and specific location, compelled by their own initiative and interest, searching for a given term or phrase.Footnote 17
Despite remaining methodological challenges to the integration and validation of search data, its advantages, especially when combined with rapidly increasing levels of Internet penetration worldwide, suggest a promising tool for understanding the actors and ideas animating society. The following section illustrates how this tool can be applied even in the highly tumultuous politics of Egypt in the aftermath of the “Arab Spring.”
Major General Mukhtar El Mallah probably deserves credit for striking the initial blow in what was to become a fiercely contested framing battle between the Supreme Council of the Armed Forces (SCAF) and demonstrators over the legacy of the “January 25 Revolution” that forced former president Hosni Mubarak from power. In September and October, as the first anniversary of the revolution loomed, numerous attempts to discredit the protestors were made, but the framing debate that began in late 2011 was qualitatively different. Speaking at a press conference on November 24, 2011, after violent demonstrations in Mohammed Mahmoud Street, General El Mallah launched a new line of attack: the protestors did not represent the will of Egypt's citizenry, most of whom wanted a return to normalcy.Footnote 18 Instead, he suggested, the protestors were provocateurs whose actions perpetuated chaos and threatened the plan for transition (specifically, the parliamentary elections which were scheduled for the end of the month). Regime apologists, pro-SCAF supporters, and individuals fed up with months of uncertainty readily embraced this phrase and its underlying logic. Soon after the press conference, slightly different permutations of the phrase were used by the military's supporters, by crowds in regular demonstrations in the Abbasiya neighborhood, and in numerous editorials, Facebook pages, and television interviews (Shadid Reference Shadid2011).Footnote 19 Some commentators subsequently interpreted parliamentary elections results as vindicating General El Mallah's characterization (Shatz Reference Shatz2012).
These anecdotes helped clarify the importance of the debate, yet it was much more challenging to assess which frame predominated Egyptian society as the anniversary of the revolution approached. Events on the ground were far too easy to interpret based on preexisting perceptions. Activists pointed to the huge and peaceful turnout on the anniversary of the revolution as definitive proof of the power and resonance of the revolution, while SCAF officials highlighted rioting after tragedy struck at a soccer match in Port Said in February as evidence that protestors were no better than thugs and opportunists.Footnote 20 Search data, gathered from four different sources are combined in this section to empirically measure these competing frames: Google Trends, which measures search trends over time; Google AdWords, which provides a 30-day rolling average of a given search term; a refined Google search for the period of December 2010 to January 31, 2012, which measures the amount of content available online; and Google Zeitgeist, which tracks the highest emerging search terms in a given year. These data cumulatively show that despite the SCAF's concerted effort to discredit ongoing demonstrations, events and figures associated with Tahrir Square and the revolution more broadly continued to captivate the attention of Egyptians online. The results from the first round of the presidential elections in May 2012 support the online evidence by demonstrating strong support for “revolutionary” candidates despite the best efforts of the SCAF to convince Egyptians that the revolution was over.
The unpopularity of President Mubarak and the largely peaceful nature of the protests meant that most Egyptians viewed demonstrators positively. The army's refusal to fire on protestors during the key 18-day period of the revolution, in stark contrast with the police, meant that Egyptians were also overwhelmingly supportive of the military, and initially the SCAF seemed to enjoy widespread support (Pew Global Attitudes, April 25, 2011). However, as street protests continued and occasional violence between demonstrators and security forces flared, evidence of polarization emerged (Younis and Younis Reference Younis and Younis2012). By October and November 2011, as confrontations between protestors and security forces became more frequent, it seemed clear that the demonstrators, an amalgam of movements with few overarching goals now that Mubarak was no longer in power, did not have a unified strategy for winning over Egyptian society. The Egyptian military, in contrast, was far more pragmatic in its approach. As the anniversary of the revolution neared, its strategy seemed to broadly focus around a single goal: convincing the Egyptian people that the revolution was over. To achieve their mission, the SCAF sought to limit attention to embarrassing incidents and discredit popular ideas and figures associated with the revolution who continued to advocate for demonstrations and the end of military rule.
Search data from Egypt in 2011 suggests that attention to events in and around Tahrir Square had not vanished as the year came to a close (table 1). Whereas the largest peak in searches directly related to events in the square occurred in late January and early February 2011, Google's Zeitgeist project that captures the fastest-rising search trends in Egypt and other countries reflects the profound influence of political developments on Egyptian search trends. Although more ambiguous as a proxy for prorevolutionary sentiment, total searches for the term “Mubarak” in 2011 exceeded annual searches for the Egyptian pop-star Amr Diab for the first time since Google started publishing normalized search data, and three different peaks for “Mubarak” in mid-February, April, and early August 2011, dwarfed even the sharp spike in searches for the hit 2011 TV series, “Forbidden Love” (figure 1).
Table 1 “Fastest Rising Searches” in Egypt, Saudi Arabia, and United States
Source: Google Zeitgeist
Figure 1 Search Trends for “Mubarak” (dark grey) and “Amr Diab” (light grey)
Source: Google Trends
Much as the SCAF might have wished otherwise, search data strongly suggested that more than a year after the revolution, events and figures associated with Tahrir Square continued to capture the attention of Egyptians well into 2012.
Google's Zeitgeist tool breaks down individual search terms into categories to facilitate more meaningful comparisons of the top-rising searches in a given year. This data painted a slightly rosier picture for the SCAF. Even considering the tremendous amounts of attention generated by events in Tahrir, the top six searches in the entertainment category, mostly related to pop stars and television shows, far exceeded the total volume of searches in the “individuals” category. High levels of search for figures associated with sports and entertainment might be interpreted as a welcome return to normalcy after a year of political turmoil and upheaval, particularly given the resonance of these figures with the youthful demographic which had been so critical to sustaining the protests.
Nonetheless, Google's Zeitgeist data for the top 10 rising searches pertaining to individuals in the news in 2011 presented further trouble for the SCAF. Not surprisingly, all of these figures were overtly connected with the revolution (table 2). Of the top 10 individual searches only Tamer Ashour, a popular singer, reflected interest in events outside of Tahrir. The other nine were all directly associated with the revolution (table 2). The surge in searches for Mubarak, Hussein Salem, Ahmed ʿAz, and Gamal Mubarak are safely interpreted as online manifestations of both the schadenfreude and shock unleashed by the revolution, as each of these previously untouchable figures was put on trial, or in the case of Salem, fled the country. The other newsworthy individuals were either connected with actual events in the square, like Ghonim and al-Sharif, or directly channeled themes from the revolution, like al-Gakh and Qatamish, competitors on an “American Idol” style program for poetry where both drew on events, themes, and figures from the revolution in their performances.Footnote 21
Table 2 Top Ten Fastest Rising Searches (People) Egypt
Source: Google Zeitgeist http://www.googlezeitgeist.com/eu
These results strongly suggest that the SCAF's attempts to portray demonstrators in Tahrir as disconnected from the rest of Egypt, as of the end of January 2012, were not effective online. As a statement of fact, General El Malah was correct to assert that protestors in the square did not “represent” all Egyptians and polling at the time suggested that many Egyptians no longer viewed protests as productive (Younis and Younis Reference Younis and Younis2011). Nonetheless, if interest in the revolution was failing or if there was no substantive connection between regular Egyptians and demonstrators in Tahrir Square, Internet searches for information about revolutionary figures and developments in Tahrir would not have been sustained at such impressive levels.Footnote 22
Even worse for the SCAF, several prominent failures, including the killings of Coptic demonstrators at Maspiro, the hub of official Egyptian media, the killing and beating of protestors in November and late December 2011, and in early February 2012 the death of approximately 70 fans at as soccer match in Port Said, caused huge surges of online attention. Starting in October of 2011, spikes in searches for the terms “al-jaysh” (the army) and “shuhada” (martyrs) were seen as well as strong levels of correlation between these two terms throughout this crucial period leading up to the first round of parliamentary elections (figure 2). Although there were subsequently substantial diversions in the trend lines, particularly in December and February when there was no direct linkage between the army and the deaths of protestors, the correlation between the search terms “army” and “martyrs” in October and November suggests that at a key moment, millions of Egyptians potentially associated the military and the deaths of innocent protestors. Even more ominously for the SCAF, search trends for the term “martyrs” cuts across the entire country (figure 3).Footnote 23
Figure 2 Search Trends for the Terms “Army” (black) and “Martyrs” (grey)
Both Boolean phrases have been adjusted to remove results related to erroneous terms e.g. “Israel” “Libya” or “Gaza.” There is a small spike in the search term “army” on October 6–7 which coincides with the National holiday commemorating the October War. The sharp spike on the October 10 coincides with the Maspiro incident. Small spikes in late November coincide with confrontations in downtown Cairo, notably Mohammed Mahmoud Street.
Figure 3 Geographic Dispersion of Searches for the Term “Martyrs” and Trend Line over Time
Darker grey on the map represents higher change in volume of searches for term in governorate.
The absence of reliable survey data from the period of interest makes it difficult to validate these results with more familiar research methods. However, as the presidential campaign got under way in the spring of 2012, SCAF's attempts to paint demonstrators as disconnected from Egyptian society, and the revolution as concluded, had clearly failed. Despite a series of legal twists and turns that disqualified two extremely popular figures, the two “revolutionary” candidates Hamdin Sibahi and Abd al-Monim al-Fatouh won a combined 40% of the popular vote (Kirkpatrick and Fahim Reference Kirkpatrick and Fahim2012). While neither candidate made it to the second round, their success strongly suggests that the SCAF had failed to convince a large segment of Egyptian society that the revolutionary figures of Tahrir Square and their fellow demonstrators “did not represent” the Egyptian people (Fahim and Stack Reference Fahim and Stack2012).
CONCLUSION
It is still far too soon to decide the legacy of the Arab Spring and the January 25 Revolution, or the sustainability of the revolution's goals in Egypt, however, search data can clearly help political scientists track broad and cross-cutting societal developments as they unfold. A final question naturally emerges from this case study: what conditions facilitate the use of search data? This analysis suggests two main factors.
Public Opinion and Internet Penetration
The empowerment of public opinion and rapidly increasing levels of Internet penetration are the twin drivers that give search data its utility for political scientists. If Internet penetration levels are too low, or there is no meaningful way for society to engage in political action, search data is only marginally useful. For example, conditions in Mauritania, Sudan, and Yemen make it highly unlikely that search data from these countries would indicate anything beyond the behavior and interests of a handful of local elites.Footnote 24
Validation
Skeptics are correct to treat online data and methods for analyzing it cautiously (Lynch Reference Lynch2011, see also Mellon Reference Mellon2013, this issue). Discrepancies between online and offline behavior do not mean that search data is somehow not “real,” but that its contribution to any given research question depends on triangulation with surveys or complementary data sources like election results, which allow a researcher to iteratively test online behavior against offline outcomes.Footnote 25 When public opinion is empowered, levels of Internet penetration are above a crucial threshold, and conventional measurements of public opinion are present political scientists should be able to use Google's “unintentional database” to explore the interests of a large section of society even in times of enormous social turmoil. Furthermore, the format and accessibility, speed of data collection, low cost, and volume of search data open new categories of research questions for exploration by social scientists—allowing researchers to quickly and efficiently test how widespread an intriguing anecdote actually is, or to objectively measure the online influence of public figures, ideas, and issues in real time.
ACKNOWLEDGMENTS
The author thanks, Sheena Chestnut-Greitens, Nelly Lahoud, Jonathan Mellon, and. Arie Perliger for their extensive comments and feedback. Thanks also to Joshua Goldstein, Najeeb Jarrar, and Ashraf Maklad for their suggestions pertaining to numerous technical and theoretical difficulties. Thanks to Jakob Earle and Muhab Wahby at the Danish Egyptian Institute for dialogue, Muhammad at Thawra Stats, and Craig Charney who all graciously explained their respective polling projects to me. Finally, several Egyptian friends weighed in at various times to keep me on track: Soha Gad, Waleed Nassar, and Noran Eid. All errors in this article are my own.
