A growing literature suggests that terrorist violence increases support for hardline nationalistic parties. Evidence of this phenomenon has been found in Israel after the Second Intifada,Footnote 1 Turkey after the PKK attacks,Footnote 2 the United States after 9/11,Footnote 3 and several European countries that experienced terrorist attacks in recent years.Footnote 4 Much of the literature argues that increased support for hardline parties reflects changed preferences in the target population. After exposure to terrorism, individuals become more sensitive to personal security and resist concessions—a policy position primarily promoted by hawkish parties.
While terrorism has strong contemporaneous effects on targeted populations, its deeper and more long-lasting legacies are not well understood. In Israel, popular support for right-wing, nationalistic parties has not diminished since the end of the Second Intifada; instead, it has remained strong—and even increased—over the years (see Figure 1). Public support for peace negotiations has similarly decreased over the last two decades (see Figure 2). New surveys of Israeli youths reveal that support for right-wing politics is greater than ever among younger generations: in 2016, over 67 percent of Jewish high-school students self-identified as right-wing supporters, and many did not regard peace negotiations with the Palestinians as an important national goal.Footnote 5
Figure 1. Rising support for right-wing parties over time
Figure 2. Decrease in public support for negotiations over time
What might explain the long-lasting popularity of right-wing ideology in Israel? What can account for the increasing support for hardline nationalistic world views in the population, even long after the Second Intifada ended and terrorist violence subsided? There are many possible explanations, ranging from elite manipulation of public opinion to demographic changes in the population. But an important insight can be found in a long history of inquiry into the roots of nationalism, which has emphasized the role of mass media in spreading and inspiring nationalist sentiment. Anderson suggested that nationalism arose in eighteenth-century Europe with the mass printing of books, which allowed new ideas about the nation-state to disseminate across communities, reproduce, and become normalized over time.Footnote 6 Posen argued that media played an essential role in motivating mass armies by facilitating the spread of a cohesive national identity in the population.Footnote 7 Other scholars have suggested that mass media can inspire violent conflict, especially when state leaders manipulate the media to stir up patriotic sentiment and out-group antagonism.Footnote 8
I make an empirical contribution to this literature by showing that right-wing, nationalistic ideology has been on the rise in tens of thousands of Israeli books since the Second Intifada. Scholars studying mass media in Israel have suggested that state actors, especially in the educational sector, have sought to influence the content of books to inspire nationalist sentiment.Footnote 9 Some even argue that the intractability of the Israeli-Palestinian conflict is rooted in the perpetuation of an atmosphere of conflict in mass media, education, and popular culture.Footnote 10 This issue has been salient in Israeli public discourse recently, with the publication of a new version of a popular high-school civics textbook that was significantly revised in 2016 to portray Israel's society and history in a way that aligns with the views of the political right.Footnote 11
Nonetheless, the argument that right-wing ideology has been disseminating in Israeli media has yet to be put to a rigorous empirical analysis. I seek to advance current research by introducing a novel text-as-data method for the large-scale analysis of political content in books, and by presenting a quantitative examination, using this tool, of right-wing content in books published in Israel between 1980 and 2008. Employing a unique data source on the content of over 70,000 published books, my study systematically analyzes the extent to which right-wing nationalistic content increased in Israeli media after periods of terrorism and over time.
The results display several interesting patterns. I find that years with high levels of terrorism are associated with a significant increase in the frequency of right-wing ideological content in tens of thousands of popular Israeli books. The increase in right-wing content does not subside, but persists over years, and is especially pronounced in books published in the latter half of the 2000s. These findings hold across multiple specifications using different measures of terrorism, and are robust to the inclusion of year fixed effects and more than a million phrase fixed effects. This striking empirical evidence closely tracks the predictions in existing literature on the central role of mass media in cultivating and sustaining nationalist sentiment, especially when considering the continuing popularity of right-wing ideology among the Israeli public since the end of the Second Intifada. Even though the findings do not necessarily imply that the content of the media has caused the Israeli public to sustain its support for right-wing politics, they nonetheless show—for the first time on a large scale, using tens of thousands of books—this systematic link.
The study therefore advances existing research on nationalism and the media by presenting a new technique to quantitatively study political content in a very large number of mass-produced books. It also contributes to research on terrorism and right-wing politics by showing that right-wing content has been disseminating in Israeli media since the Second Intifada, a finding that closely tracks similar changes in public opinion over the years.
Text as Data: Studying the Political Content in Thousands of Books
I use automated text analysis to quantitatively measure right-wing content in books over time. Since the goal is to uncover systematic patterns in a large number of books, I use a unique data set, the Google N-gram, which contains granular information on the content of tens of thousands of books published in Israel over more than two decades.Footnote 12 The content of books is analyzed using two-word phrases as the unit of analysis. I examine how the usage of specific phrases reflecting right-wing ideology changed over time and after periods of terrorism.
The Google N-gram data set was created from a corpus of digitized books in eight languages, including English, French, Russian, and Hebrew, among others, drawn from several dozen universities around the world and direct contributions by publishers. The latest version of the data set consists of more than 8 million volumes, which is about 6 percent of books ever published. The textual information was collected by scanning books and digitizing their content with optical character recognition, and includes metadata on the year and place of publication.Footnote 13 In light of copyright restrictions, the data set does not provide the content of books in full-text format, but consists of a matrix of n-gram frequencies by year. An n-gram is a sequence of several words. Google N-gram includes phrases of different lengths, up to five-word phrases, and is limited to phrases that appear forty times or more in the corpus.Footnote 14 This study uses the “2-gram” (i.e., two-word phrases) version of the Google N-gram corpus in Hebrew, which consists of more than 70,000 books and over 8 billion words.Footnote 15 A brief summary of the data collection process follows (see the supplementary appendix for a detailed description).
Creating a Vocabulary of Right-wing Phrases
I employ two textual sources to find words that are linked to right-wing ideology. First, I use Israeli political party platforms from elections taking place between 1981 and 2013 to generate a “core vocabulary” that represents political issues on the right-left ideological space.Footnote 16 This corpus consists of fifty-one platforms from right-wing, center, and left-wing parties. Table A1 in the supplementary appendix lists the parties that competed in elections from 1981 to 2013, divided by blocs. Second, I use the full text of the Hebrew edition of Wikipedia, which consists of 156,531 articles, to expand the core vocabulary. Since the vocabulary of party platforms is limited to words that parties use in their official documents, it is likely to not fully capture the range of words used in popular books to describe the same political ideas. To better measure the usage of political phrases in books, I expand the party platform vocabulary by adding associated terms from Wikipedia. I divide each Wikipedia article into paragraphs and calculate the paragraph-level association between core phrases coming from party platforms and associated phrases appearing in Wikipedia. The association reflects the number of co-occurrences of each Wikipedia phrase with core vocabulary phrases, with larger number of co-occurrences reflecting a higher association.Footnote 17
Figure 3 shows a random sample of phrases that are most strongly linked to the political ideology of parties from the right and left blocs. All phrases are my own translations from Hebrew to English. Although there is some noise, the overall pattern shows that phrases linked to right-wing parties tend to focus on Jewish nationalism, Jewish land, and wars, for example, “Jewish people,” “conquest land,” and “after war.” On the other hand, phrases linked to left-wing parties relate to topics such as civil rights, Arab citizens of Israel, and peace, for example, “situation of rights,” “Arab population,” and “just peace.” While these phrases, on their own and in the abstract, do not necessarily imply right- or left-wing ideology, their usage in Israeli discourse in Hebrew signals different political slants. Thus, for example, the phrase “Jewish people” is frequently used in the context of Jewish nationalism, while the phrase “just peace” is a commonly used by the left to discuss the resolution of the Israeli-Palestinian conflict.
Figure 3. Phrases most strongly linked the ideology of political blocs
This distribution of phrases generally aligns with what is usually conceived of as right and left in Israeli politics. While policy preferences are multidimensional, much of the divide between the right and the left in Israel revolves around the Israeli-Palestinian conflict.Footnote 18 In general, the left believes that the best way to end the conflict is through peace negotiations and concessions, while the right is skeptical about the success of concessions, stressing the importance of military force for defense and deterrence.
To illustrate how these ideological phrases might appear in books, Figure 4 presents an example of sentences from books published in Israel that include phrases linked to right-wing and left-wing parties. These sentences were originally written in Hebrew—I translated them to English. Phrases linked to right-wing parties, such as “Jewish people” and “after war” are located in sentences that discuss the value of territory for the Jewish people and the importance of security for the state of Israel. These issues reflect the ideology of the right-wing bloc, which tends to prioritize territory and national security. On the other hand, phrases that are linked to left-wing parties, such as “situation of rights” and “just peace,” are found in sentences that discuss human rights violations and peace negotiations—issues advanced by left-wing parties.
Figure 4. Examples of phrases in books
Measuring the Usage of Right-wing Phrases in Books
I measure the yearly frequency of phrases linked to right-wing ideology in the Google N-gram corpus. The yearly frequency of each phrase reflects the number of times in which it appears in new books published each year. I also measure the frequency of phrases reflecting left-wing and centrist ideology, for comparison purposes. The frequency is translated to a percentage of the sum of the frequencies of all phrases in my vocabulary. This takes into account the fact that not all phrases appearing in Google N-gram each year are in the vocabulary of political phrases that I generated. Controlling for this possibility ensures that changes in frequencies of phrases outside my data set do not lead to artificial changes in the frequencies of phrases in my data set.Footnote 19
Data on Terrorism in the Second Intifada
Since a large literature has suggested that rising support for the right in Israel is driven by terrorism in the Second Intifada,Footnote 20 and that sustained support for nationalist ideology is likely to be enhanced by its dissemination in mass media,Footnote 21 I examine whether there is a systematic relationship between over-time changes in right-wing content in books and terrorist violence in Israel. To do so, I match the frequency of phrases found in books to two aggregate measures of terrorist violence, both based on the number of casualties during the conflict. The first compares the content of books before and after terrorism peaked in the Second Intifada in the year 2002 (see Figure 6). I compare the frequency of phrases reflecting the ideology of right-wing parties before and after this peak by carrying out various estimations with different year intervals. The first estimation measures the usage of right-wing phrases six years before and six years after the peak of terrorism.
In the second estimation, I observe changes in phrase frequency using a greater distance from the peak, where I utilize six years before and after the peak of terrorism but exclude the three years on either side of the peak. The third estimation narrows the before and after window to three years. In interval notation, the first estimation measures [–6, –1] and [1, 6] years from the peak, the second estimation measures [–6, –4] and [4, 6] years from the peak, and the third estimation measures [–3, –1] and [1, 3] years from the peak. Measured in this way, all estimations allow for at least a one-year difference between the peak of violence and the year in which books are published.Footnote 22
The second measure uses the yearly number of casualties during the conflict. Data on Israeli casualties come from B'Tselem, the Israeli Information Center for Human Rights in the Occupied Territories.Footnote 23 First, I calculate the cumulative number of casualties in the Intifada to understand whether the frequency of hawkish ideology increases with the number of victims from terrorist violence. Second, I create five-year lags and moving averages of the number of casualties to estimate patterns over time. Finally, I break down the types of casualties between civilians and soldiers to examine whether the frequency of right-wing ideology in popular books changes differently between combatant and noncombatant casualties.
This data collection process resulted in an original, phrase-level data set on the political content of books over time. This data set includes over a million phrases whose frequency is measured over the course of twenty-eight years, from 1980 to 2008. Each phrase is linked in probability to political parties, measured as a weight ranging from 0 to 1.Footnote 24 Figures 5 and 6 present summary statistics for the entire data set by year.
Figure 5. Summary statistics: Phrase frequency over time
Figure 6. Israeli casualties in the intifadas
Estimation
I examine the relationship between terrorism and right-wing content in books using several ordinary least squares models. First, I estimate a model that compares the frequency of phrases linked to right-wing, left-wing, and centrist parties before and after the peak of violence in the Second Intifada:
$$y_{i,t} = \beta _1(R_i \times T_{i,t,w}) + \beta _2(C_i \times T_{i,t,w}) + \beta _3(L_i \times T_{i,t,w}) + \gamma _1R_i + \gamma _2C_i + \gamma _3L_i + \alpha _i + \delta _t + \varepsilon _{i,t}$$In equation 1, y i,t represents the frequency of phrase i at year t in Google N-gram; R i, C i, and L i are indicator variables coded 1 when a phrase is linked to a party from the right, center, and left-wing blocs, respectively, and 0 otherwise; T i,t,w is an indicator variable coded 1 when phrase i at year t appears after the peak of terrorism and 0 if before; and the subscript w indicates the size of the before and after window around the terrorism peak in the Second Intifada. Finally, α i and δ t correspond to phrase and year fixed effects, and ε i,t is the error term.
The second set of models is based on the same structure, but examines the relationship between the number of casualties from terrorist violence and the usage of right, center, and left-wing phrases in subsequent years:
$$y_{i,t} = \beta _1(R_i \times K_{i,t,\tau} ) + \beta _2(C_i \times K_{i,t,\tau} ) + \beta _3(L_i \times K_{i,t,\tau} ) + \gamma _1R_i + \gamma _2C_i + \gamma _3L_i + \alpha _i + \delta _t + \varepsilon _{i,t}$$Here, K i,t,τ captures the number of people killed in terrorist attacks in time period τ ≠ t, which is measured differently in each model. First, I estimate the cumulative number of casualties in the Second Intifada. In this estimation, each additional year increases the number of individuals killed. The second set of estimations use five-year lags and moving averages of the total number of casualties. Finally, I estimate lags and moving averages where the number of casualties is broken down between civilians and soldiers.
In all models, the main coefficient of interest is β 1, which estimates the relationship between terrorism and the frequency of the use of phrases related to right-wing ideology. All regressions are estimated with probability weights to account for phrases’ proximity to the ideology of political parties.
The use of phrase and year fixed effects is an important part of my estimation strategy for several reasons. First, it enables me to analyze the relationship between terrorist violence and right-wing content in books more cleanly, by making comparisons within phrases over time, taking into account yearly trends in phrase frequencies. Second, it allows for a more accurate estimation of the standard errors. One of the drawbacks of using time-series data in difference-in-difference models is that unmodeled mechanical trends in the dependent variable can bias standard errors downward because of serial correlation in the error term.Footnote 25
While language is trended in some sense, especially when certain topics become more popular over time, any variation in the frequency of phrases that remains after controlling for time trends, differences across phrases, and terrorist events that might affect the use of language, is likely to be independent and identically distributed (i.i.d) white noise. The year fixed effects absorb linear trends in the linguistic popularity of all phrases in a given year; the phrase fixed effects subtract out time-invariant variation in the frequency across phrases, and the variable capturing terrorist violence accounts for theoretically substantive variation in phrase frequency over time.
In addition, the yearly frequency of phrases reflects the number of times in which phrases appear in new published books each year. These books are written by many different independent authors in various types of books. For example, in the year 2003, there were 7,128 new books published in Israel, 82 percent of them were published by commercial, nongovernmental, or private publishers. In 2007, 5,850 new books were published, of which 86 percent were printed by commercial, nongovernmental, or private publishers.Footnote 26 Thus it is plausible to assume that any residual variation in the frequency of phrases not accounted for by the model is not systematically trended. The supplementary appendix provides statistical tests showing that residual variation is uncorrelated over time.
Results
Figure 7 reports the main results. The left panel shows results from the model in equation 1, which uses the peak of terrorism as the predictor. All points in this panel present the coefficient and 95 percent confidence intervals for the interaction term: After peak of terrorism × Right, which captures the change in the frequency of phrases linked to the ideology of right-wing parties after the peak of terrorism. The numbers next to the points present the percent change in the frequency of right-wing phrases associated with each coefficient. The left-most point presents the coefficient obtained when measuring phrase frequency [–6, –1] and [1, 6] years from the peak of terrorist violence; the middle point plots the coefficient when phrase frequency is measured [–6, –4] and [4, 6] years from the peak of violence; and the right-most point reports the coefficient estimated using measures of phrase frequency [–3, –1] and [1, 3] years from the peak.
Figure 7. Terrorism and right-wing content in books
The right panel reports the results from the model in equation 2, which uses the number of casualties to measure terrorism. Here, all points represent the coefficient on the interaction Number of casualties × Right, along with 95 percent confidence intervals. The left-most point shows the coefficient estimated when using the cumulative number of casualties in the Second Intifada as the predictor; the middle point shows the result obtained when using a five-year moving average of the number of casualties; and the right-most point shows the coefficient estimated when using a five-year lag of the number of casualties. In these models, the relationship is measured as an increase of one casualty so the coefficients are notably smaller than in the left panel, which measures differences across several years. In all estimations, results are calculated with phrase and year fixed effects. In addition, all coefficients are multiplied by 10,000,000 for easier presentation, because the phrase-level unit of analysis renders the magnitude of the original coefficients very small. The full regression results are reported in Tables 1 and 2.
Table 1. Frequency of phrases by bloc after the peak of terrorism in the Second Intifada
Notes: The table reports changes in the frequency of phrases linked to the ideology of political blocs in Israel after the peak of terrorism in the Second Intifada (2002). The association between the peak of terrorist attacks and frequency of right-wing, left-wing, and center phrases is captured by the interaction terms. All coefficients are multiplied by 10,000,000 for easier interpretation. As analysis is conducted on the phrase level, the magnitudes of the original coefficients are very small. *p < .10; **p < .05; ***p < .01.
Table 2. Frequency of phrases by bloc and the number of casualties from terrorist violence
Notes: The table reports the relationship between the various specifications of the number of casualties from terrorism in the Second Intifada and the frequency of phrases reflecting the ideology of political blocs in Israel. The association between the number of casualties and frequency of right-wing, left-wing, and center phrases is captured by the interaction terms. All coefficients are multiplied by 10,000,000 for easier interpretation. As analysis is conducted on the phrase level, the magnitudes of the original coefficients are very small. *p < .10; **p < .05; ***p < .01.
As Figure 7 presents, the results show a consistent pattern whereby phrases linked to right-wing parties increase in popular books in years succeeding periods of terrorism in Israel. When comparing the content of books [–6, –1] and [1, 6] years or [–6, –4] and [4, 6] years from the peak of terrorism, we see that the frequency of phrases reflecting the ideology of right-wing parties significantly increases. This change is substantively significant: the frequency of phrases reflecting hawkish ideology is about 2 to 6 percent higher in books published after the peak of the Second Intifada, compared to their frequency in pre-Intifada years. Given that any given phrase can only appear a limited number of times in a document, this increase is notable. This relationship does not hold when comparing phrase frequency [–3, –1] and [1, 3] from the peak of terrorism. One explanation is that it takes more than three years for book content to change, especially considering the time it takes to develop an idea for a book, write a draft, and undergo the process of review, revision, and publication.
Looking at the results using the number of casualties as the predictor, we find a similar pattern. Phrases linked to right-wing content significantly increase with the number of casualties from terrorist violence. In substantive terms, each person killed is associated with a .02–.04 percent increase in the frequency of right-wing phrases. While the percent change seems small, it should be noted that it reflects the change in the frequency of one phrase that is associated with one casualty. In the supplementary appendix I report additional estimations, showing that the results hold also when breaking down the casualties between civilians and combatants, as well as when using measures of terrorist violence from the First Intifada.
Comparing these findings to the results obtained for phrases linked to left-wing and centrist parties (see Tables 1 and 2 and additional results in the supplementary appendix), I find that left-wing phrases either do not change, or even significantly decrease after the peak of terrorism in some models. To recall, phrases reflecting left-wing ideology are related to topics such as civil rights, Arab citizens of Israel, and peace negotiations. Phrases linked to centrist parties significantly increase in books after periods of terrorism, a result which holds in almost all model specifications. It is hard to evaluate what types of issues are captured by centrist phrases because these phrases come from a noisy vocabulary mixing right-wing and left-wing positions (see Figure A2 in the supplementary appendix).
Finally, I examine the sensitivity of my results to the choice of arbitrary placebo “peaks” in terrorist violence by estimating the model described in equation 1 for each year in my data set. The choice of the peak of violence in my main estimations is based on the notion that the year with the highest number of casualties (the year 2002) can plausibly capture the relationship between terrorism and the usage of right-wing content in popular books. However, it is equally likely that other years during the Intifada also capture this association. If terrorist violence has a systematic relationship with the usage of right-wing content in mass media, we would expect other Intifada years to have a positive relationship with the frequency of right-wing content in books in subsequent years. Furthermore, this logic implies that arbitrary “peak” years outside of the Intifadas should not show an increase in the usage of right-wing phrases.
Figure 8 shows the results. The figure plots, for each year, the coefficient on the interaction term After peak of terrorism × Right, where the before and after windows around the peak are set to six years. For each year, the point represents the change in the frequency of phrases linked to right-wing parties in the six subsequent years, compared to their “baseline” frequency in the six prior years. The figure also plots the yearly number of casualties from terrorist violence, and shades in gray the Intifada years. Overall, there is a systematic increase in the usage of phrases linked to right-wing ideology in books published in years following the Intifada years. These results are stronger for the Second Intifada, especially for phrases appearing in books published in the second half of the 2000s. The figure also shows that non-Intifada years are not systematically related to an increase in the right-wing content of books.
Figure 8. Placebo: Usage of right-wing phrases in different specifications of “peak” years
Overall, the results show that terrorism is associated with a systematic increase in the frequency of content relating to right-wing ideology in Israeli books, an increase that has become more pronounced over time. These findings hold in tens of thousands of books, using different measures of terrorist violence, and are robust to various model specifications and a battery of robustness tests. Because the Google N-gram data set ends in 2008, it is not possible to say how the dissemination of right-wing content would have looked in the 2010s, but qualitative evidence from a recently published educational textbook suggests that this trend is continuing.Footnote 27 These empirical findings support the argument that mass media have played an important role in promoting and sustaining long-lasting right-wing nationalism in Israel.
Possible Explanations for the Rise of Right-wing Content in Books
Changes in Popular Demand for Right-wing Content
One explanation for the rise in right-wing content in Israeli books is that profit-maximizing book sellers respond to changes in popular demand by consumers. A growing number of studies from Israel have found that terrorism in the Second Intifada increased public support for right-wing parties,Footnote 28 a trend that has also been documented in Israeli national elections and public opinion polls.Footnote 29 After the Second Intifada, public support for right-wing politics has been growing, especially among younger generations.Footnote 30 Thus, it is possible that book publishers have been considering these shifts in population preferences when choosing content for publication.
This notion is consistent with existing research on mass media markets, which views media as a commercial enterprise producing works to maximize profits.Footnote 31 It also aligns with Anderson's argument in the context of European nationalism that print capitalism gave book publishers incentives to sell books on topics that interested the masses. Genzkow and Shapiro's finding that political slant in American newspapers is driven by consumer demand also supports this notion.Footnote 32 Thus, it is possible that over-time changes in public demand for right-wing content, especially since the end of the Second Intifada, may have led to increased production of such content in the Israeli market for popular books.
Elite Manipulation of Mass Media Content
The causality may also run in reverse: ongoing public support for right-wing politics in Israel could have been facilitated by the dissemination of right-wing content in Israeli mass media. Under this view, political elites are strategically influencing what is published to inspire nationalist sentiment in the population. Qualitative evidence from Israeli high-school textbooks suggests that elite manipulation of content has been taking place for decades.Footnote 33 Recent examples include the banning in 2015 of a popular book narrating a love story between a Jewish Israeli and a Muslim Palestinian on the grounds that it threatens Jewish nationalist identity,Footnote 34 and a series of significant revisions to a 2016 edition of an Israeli civics textbook that more strongly reflect the views of the political right.Footnote 35 Scholars have argued that a conflict atmosphere in Israeli mass media has contributed to the perpetuation of the Israeli-Palestinian conflict.Footnote 36
Snyder and Ballentine discuss the ability to strategically increase nationalistic content in the media as a market failure in the “marketplace of ideas.”Footnote 37 Several studies have shown that this can happen in the context of terrorism when elites take advantage of the “rally around the flag” effect and strategically influence the content of the media to sway the public to support hawkish policies.Footnote 38 The United States after 9/11 presents a good example of this phenomenon. The Bush administration was able to persuade the public to support the 2003 Iraq invasion by manipulating the content of the news.Footnote 39
However, it is unclear how much control elites have over the publishing process of popular, non-educational books. Unlike school curricula, or even news media, which have been shown to be heavily influenced by those with political power,Footnote 40 popular book publishing is a slow-moving and diffuse process that is harder to systematically manipulate. The supplementary appendix describes an empirical exercise evaluating the extent to which political elites in Israel have influence over book publishers’ choice of content. The evidence reveals that about 10 percent of the publishers are subject to possible governmental influence. This suggests that changes in the right-wing content of books cannot be completely dictated by political elites.
Demographic Changes in Israeli Society
A third explanation relates to deeper structural changes in Israel's demographics, specifically, waves of immigration from the former Soviet Union and the rapid population growth in Jewish settler communities. Since the 1990s, over a million immigrants from the former Soviet Union immigrated to Israel, a significant proportion compared to the total population.Footnote 41 The arrival of Russian immigrants strengthened Israel's political right because immigrants have been strongly favoring right-wing, hardline policy positions, especially with respect to the Israeli-Palestinian conflict.Footnote 42 Over the same time period, the Jewish settler community also experienced a rapid population growth. In 2015, the annual growth rate of the Jewish settler population was 4.1 percent, which was twice the growth rate of the general Israeli population.Footnote 43 Jewish settlers tend to support hardline, nationalistic policy positions and constitute a core group of voters of right-wing parties in Israel.Footnote 44
These demographic shifts in the composition of the Israeli population might have also contributed to the increase in the right-wing content in Israeli mass media. The mechanisms could be related to changes in demand, as a larger share of the population favors right-wing ideological content; or, alternatively, increased political leverage by elites to influence media content. While the process driving these changes is complex and multidimensional, the pattern I found is powerful and robust—in more than 70,000 books published in Israel over the course of twenty-eight years, content related to right-wing nationalism has become more frequent since the end of the Second Intifada.
Conclusion
This research note provides systematic empirical evidence that right-wing content has been increasingly disseminating in Israeli popular books over the years. These findings support a large body of work on the roots of nationalism, which has long emphasized the role of mass media in cultivating nationalist sentiment. It also contributes to existing work on the legacies of terrorism, which has so far mainly focused on contemporaneous changes in voting behavior by showing that terrorism, at least in the Israeli case, has had a much deeper and longer-lasting impact, sparking long-term changes in the content of mass media.
Another contribution of this study relates to its use of automated text analysis methods. A growing number of studies in various social science disciplines use automated methods to analyze the content of large-scale textual sources. But there is currently very little political science research employing automated text analysis in Hebrew. This study introduces an algorithm to morphologically stem Hebrew words—an algorithm that has not been previously utilized on Hebrew text in the political science literature. As such, it contributes to existing work on automated text analysis by providing tools to pre-process Hebrew-language texts. This enables one to examine the way in which terrorism influences the content of a large number of Israeli books, evidence that is impossible to observe without such methods.
Future research could apply this method to other countries that experienced terrorism and/or a rise in right-wing nationalism in recent years. Google N-gram currently contains data on hundreds of thousands of books in English, French, German, Italian, Russian, Spanish, and Chinese. Further studies could also examine whether the pattern I found holds in other types of media outlets, such as newspapers, blogs, and social media. Terrorist violence has been shown to have a substantial impact on targeted populations, and further research is needed to better understand its consequences.
Supplementary Material
Supplementary material for this article is available at <https://doi.org/10.1017/S0020818318000383>.
