The ground operation sent citizens into a frenzy: the rally around the flag effect during operation protective edge

ABSTRACT

This study examines the rally round the flag effect during Operation Protective Edge, a war between Israel and Palestinians. Research has demonstrated the mechanisms by which military crises escalate, and the public perceives the threat, which in turn leads to support for government leadership. Forming the basis of this theory is Social Identity Theory; however, in the context of international relations, there seems to be little research that examines the mechanism from which Social Identity Theory causes the rally to affect the minority group. This is probably because, in inter-state and inter-ethnic conflicts, there have been few opportunities to access research information on situations where there is a part of a hostile nation or ethnic group within one state. Public opinion polls were conducted repeatedly during Operation Protective Edge, and Arab Israelis were included in the target population for this poll. Subsequently, the mechanisms that determined the attitudes of Arab Israelis during this Israeli-Palestinian military conflict can be determined. This study adjusted for covariates by using an Interrupted Time-Series analysis to infer causality of the ground war rush on the rally effect.

KEYWORDS:

• Arab-Israeli conflict
• rally effect
• social identity theory
• interrupted time-series analysis

Introduction

Coverage of the start of a battle or the outbreak of war often inflames and inspires people to support their government more strongly. This phenomenon, referred to as the rally around the flag effect or rally effect, begs the question of why the majority of a country’s citizens support a government’s ‘seemingly irrational’ decisions to start wars. Research on the effects of rallying around the flag has accumulated primarily in the United States, but in recent years, research from the United Kingdom, (Feinstein, 2018; Lai & Reiter, 2005) Israel, (Hamanaka, 2018) and Japan (Kobayashi & Katagiri, 2018) has emerged. This is unsurprising, as the rally around the flag effect is not a phenomenon unique to the United States – a country that occupies a unique position in contemporary international politics – but is a universal phenomenon.

This study examines the effects of the rally around the flag during Operation Protective Edge – the war between Israeli and Palestinian armed groups that lasted for 50 days from July 8 to 26 August 2014. Since the complete withdrawal of IDF forces from the Gaza Strip (2005), the Israeli-Palestinian conflict has often intensified, leading to large-scale conflicts such as Operation Summer Rain (2006), Operation Cast Lead (2008–09), and Operation Pillar of Defence (2012).

Operation Protective Edge caused more deaths than Operation Cast Lead, which was said to be the worst conflict on the Palestinian side (AFP BB News, 2014). According to the Palestinian Authority Health Ministry in the Gaza Strip, 2,145 Palestinians were killed during Operation Protective Edge, including 578 children, 263 women, and 102 older people. The number of injured people has reportedly exceeded 11,100, nearly a third of whom are children. (International Middle East Media Center, 2019) Conversely, on the Israeli side, 66 IDF soldiers were killed (Israel Ministry of Foreign Affairs, 2019), and only six civilians were killed by rockets and shells fired by Palestinian militants (United Nations Office for the Coordination of Humanitarian Affairs (OCHA), 2015). Subsequently, this war has been labelled as extremely asymmetric from looking at the death toll on both sides (Dekel, 2014).

Additionally, public opinion polls have been conducted repeatedly during this border defence operation. Conveniently for this study, the poll included questions that would be useful in the study of the rally around the flag effect. In addition, some Arabs with Israeli citizenship were also the subjects of the opinion poll. This was the motivation for this study, as the data are effective for identifying the conditions necessary for the activation of the rally effect.

Subsequently, the primary research question for this study was what attitudes do Arab Israelis, a minority in Israeli society, form towards the wartime government of Israel and their Arab Israeli community? This study does not deal with the normal rally effect but with an ‘inverse rally effect’ in which Israeli-Arabs identify more with the Palestinian society or are hostile to the government of Israel. As discussed below, in the context of Israeli politics, Arabs are more likely to be seen by Jews as an ‘enemy group.’ In this context, through the research question above, this study focused on how the collective psychology of minorities behave towardsthe government and their community in a hostile situation involving the majority against the minority.

To elucidate this question, the discussion will proceed in the following order. First, the following section examines social identity theory, which is the foundation theory of the rally effect, to derive a working hypothesis. Section 3 introduces the dataset used in the analysis and discusses the research design necessary for causal inference. Section 4 presents the analysis results, and, finally, we discuss the implications of the estimated causal effects.

Theory

Many international relations researchers have characterised Mueller’s research (Mueller, 1971, 1973) as a landmark in empirical studies of the public opinion on war (Baum & Groeling, 2010; Berinsky, 2009; Berinsky & Druckman, 2007; Brody, 1991; Gartner & Segura, 1998; Gelpi et al., 2006, 2009). Mueller argued that every conflict would lead to the increasing popularity of leaders and a rallying around the flag effect, proclaiming that wartime governments enjoy majority public support at the start of wars. However, Baker and O’Neal rejected a condition of the theory and argued that all military crises do not always lead to the rally effect (Baker & Oneal, 2016). According to a Haaretz poll conducted on 9 July 2014, shortly after the beginning of the Operation Protective Edge war, Prime Minister Netanyahu’s approval rating was only 40%, down four points from his approval rating a week earlier (see Figure 1).

Much of the literature supports the mechanism by which military crises escalate, and the public perceives the threat, which, in turn, leads to public support of government leaders. Theories in which threat perceptions are linked to leader support include Nation-State chauvinism and Patriotism Perspective (Feinstein, 2018). According to studies by Lambert, who elucidates the causal mechanism of the rally from the standpoint of social psychology, social identity theory (SIT) forms the basis of the rally effect (Lambert et al., 2010, 2011). SIT is based on a minimal group experiment conducted by Tajfel and Turner (Henry. & Turner, 1986). Even if a group is artificially created in an experiment, people assigned to the group demonstrate ‘in-group favoritism’ by offering favours to the group members. This behaviour is known to be highly reproducible, and the manifestation of nationalism and chauvinistic behaviour is believed to be based on ‘in-group favoritism.’ As the rally round the flag effect is also a manifestation of this, it is assumed to be based on SIT (Groeling & Baum, 2008).¹

In the context of international relations, few previous studies have examined the causal mechanism underlying the SIT and the rally round the flag effect. This is likely because, for inter-state and inter-ethnic conflicts, there have been few opportunities to access information on situations involving part of a hostile state or ethnic group within one state or ethnic group, that is, hostile subgroups within a society. The presence of a perceived hostile subgroup in society, such as Japanese Americans living in the United States during World War II or Iraqis living in Iran during the Iran-Iraq War, is not uncommon. However, there seems to be little empirical research on whether, in the context of a national security emergency such as war, enemy subgroups identify with the nation in which they reside, or with enemy countries and enemy groups that share their language and culture.

As mentioned above, studies on the rally effect are not limited to the United States, and there have been several studies on the Israeli-Palestinian conflict as a case study (Canetti et al., 2017, 2018; Gvirsman et al., 2016; Halperin et al., 2013; Hamanaka, 2018; Zipris et al., 2019). However, most of these studies were based on laboratory surveys or survey experiments and did not include analysed polling data conducted during the ongoing war. The primary interest of these prior studies is also the long-term effects of the rally round the flag effect, such as whether the experience of conflict perpetuates support for war and the military.

Among the previous research, Feinstein is an outlier as he uses experimental data conducted during an ongoing war. Similar to our study, his study uses Operation Protective Edge as a case study to elucidate the causal mechanisms by which Jewish citizens, faced with security threats, are attracted to a collective identity, and how these heightened identities stimulate in-group bias, which leads to patriotism. In other words, Feinstein’s study is also distinctive and epochal in that it directly examines the effects of SIT. By contrast, our study differs as it uses data from Arab citizens as well as Jewish citizens and examines what perceptions both Jewish and Arab citizens have of each other. In other words, do Jewish citizens see Arab citizens, who are also citizens of Israel, as an in-group, or do they see them as an out-group that secretly communicates with the enemy? Or do they come to see them as a definite out-group under the stimulus of war? Do Arab citizens consider Israeli Jews to be an in-group or an out-group? Subsequently, the additional originality of this study is the analysis of these causal relationships.

Past opinion polls have demonstrated that Arab citizens in Israel identified as Arab or Palestinian, and not as Israelis; the surveys conducted in 2000 by Eliezer Ben-Rafael and Yochanan Peres revealed that about half of Arab respondents answered that their collective identity was Arab, and almost a quarter of them said their collective identity was Palestinian. Only slightly more than a quarter answered that they were Israeli citizens (Ben-Rafael & Peres, 2005, p. 170). The 2010 Brookings poll for Arab Israelis indicated 44% of respondents identified as Palestinian compared to 33% who said they were Israeli citizens (Radai et al., 2015). Rafael Israeli, a researcher at Hebrew University, cited the Institute for the Study of Peace survey results and reported the level of Israeli Arab identification with the second intifada skyrocketed (Israeli, 2008, p. 79).

When witnessing the Israeli-Palestinian military conflict, people evaluate political leaders through political parties.² Prior research has demonstrated that Jewish Israelis support the Israeli government for carrying out war. In response, we can theoretically expect Arab Israelis to react in the following ways. If the identity of Arab Israelis is determined by their citizenship, they will be likely to support the Israeli government during the war. However, if the identity of Arab Israelis is determined by their own ethnicity, they will not be likely to support the government leading the war. Therefore, the operational hypotheses to be examined in this study are as follows:

H1a: Arab Israelis hold Israeli governments that carry out war in the same high regard as Jewish Israelis (Citizenship Identity Hypothesis).

H1b: Arab Israelis, unlike Jewish Israelis, have a low opinion of Israeli governments that carry out war (Ethnic Identity Hypothesis).

The logic of H1a is that even though a citizen is of Arab origin, if they have Israeli citizenship, they will identify with Israeli society and support the government against the threat of war. This is a simple hypothesis that links threat perception and the rally round the flag effect (Hamanaka, 2019a).³ The logic of H1b, however, is that if a citizen is Arab in origin, then their ethnic identity is linked to the Palestinian community, even if they have Israeli citizenship. Subsequently, they will not support the governments against the threat of war. In this hypothesis, the SIT links Arab Israelis and Palestinian society in the occupied territories as a unitary group, and the Israeli government is regarded as the commander of the out-group by Arab Israelis.

Assuming that Israeli society is made up of communities divided into Jewish and Arab communities, in the event of war, individuals may highly value the behaviour of individuals belonging to communities that they view as in-groups, while those belonging to communities that they view as out-groups may lowly value the behaviour of individuals belonging to communities that they view as out-groups. In other words, this involves the psychological dynamics of empathy and cohesion for the inner group and dislike and alienation for the outer group. In the form of a hypothesis, it would be as follows:

H2a: During war, Jewish Israelis evaluate the Jewish population highly and the Arab population lowly (Ethnic Identity Hypothesis).

H2b: During war, Arab Israelis hold Arabs in high regard and Jews in low regard (Ethnic Identity Hypothesis).

Although H2a and H2b have an inverted structure of subject and object, it is not necessarily self-evident that majority and minority ethnic groups react in the same manner. Just because an individual belonging to the majority in a society can express a critical attitude towards the minority community, this does not mean that an individual belonging to the minority can criticise the majority community. This is where asymmetries or power dynamics between communities may emerge.

Figure 1 illustrates the changes in the Prime Minister’s approval rating from July 3 to 27 August 2014; Operation Protective Edge was issued on July 8, but as mentioned earlier, the approval rating on July 9 was 40%, slightly lower than the previous week (44%). It was not until the early hours of July 17 that the ground war broke out, and when the Prime Minister’s approval rating increased significantly. On that day, his approval rating rose to 57%, while in the July 20 and July 24 surveys, his approval rating rose to 73% and 82%, respectively, indicating a significant surge. In other words, the results of the ground battles were reported and led to the rally effect around the prime minister. The fighting from the start of the war to the beginning of the ground war was dominated by airstrikes, during which the Palestinian militants cannot be neutralised as they can hide in bunkers. The failure of the airstrikes to neutralise the militants is indicated in the bottom panel of Figure 2, which illustrates the persistence of rocket attacks from the Gaza Strip. It is presumed that the public’s expectation that the conflict would end in a way that suppressed the armed organisations once ground warfare with direct troops was launched caused the rally effect. These, together with the previous assumptions, lead to the following hypotheses:

H3a: Jewish Israelis raise their assessment of the Israeli government from decisions to execute ground operations (Rally Effect).

H3b: Arab Israelis lower their assessment of the Israeli government from decisions to execute ground operations (Inverse Rally).

Table 1 illustrates the wording of the questions to measure Jewish and Arab evaluations of the government as the outcomes of the rally effect.

Table 1. Variables and wording of questions

Variables		Wordings of Questions
Rally round the flag
Evaluation about the government		What grade would you give to the functioning the government in running Operation Protective Edge so far?
		[1 = Very Poor … 10 = Very Good]
Evaluation about Jews		What grade would you give to the Jewish population in Israel for its behaviour since the operation began?
		[1 = Very Poor … 10 = Very Good]
Evaluation about Arabs		What grade would you give to the Arab population in Israel for its behaviour since the operation began?
		[1 = Very Poor … 10 = Very Good]

Source:The Peace Index Survey B and C

Data sets and research design

Data sets

The direct impetus for the implementation of Operation Protective Edge was the Israeli retaliation for the kidnapping and murder of three Jewish settler boys in June 2014 (Berti, 2014). However, there is a view that the war was waged by Hamas, which effectively controls the Gaza Strip, in an attempt to break the deadlock (Kyodo News, 2014). The same Muslim Brotherhood-affiliated Free and Fair Party government was overthrown in a July 2013 coup d’état, resulting in the loss of good relations between Hamas and Egypt. After the coup d’état, the Sisi regime destroyed the underground smuggling tunnels on the border between the Gaza Strip and Egypt, leaving Gaza’s economy in dire straits.⁴

The Israeli government described the purpose of the ground operations as the destruction of underground tunnels used in terrorist attacks. Civilian deaths in the Gaza Strip skyrocketed as ground troops rushed into the Gaza Strip. The Tokyo Shimbun, quoting an AFP report, said that the number of Palestinian casualties exceeded 500 from the start of Operation Protective Edge on July 8 until 20 July 2014 (morning edition, 22 July 2014). According to the New York Times, 247 Palestinians were killed between the start of the ground war on July 17 and the early hours of July 20, bringing the number of deaths in four days to the same number as in the nine days before the ground war (Yourish & Keller, 2014).

Haaretz posted a graph (top panel of Figure 2) illustrating the change in the number of deaths in the Gaza Strip (Haaretz, 2014). Figure 2 indicates that the death toll, which did not exceed 50 per day at the time of the airstrikes, rose to more than 50 each day after the ground war broke out. Simultaneously, the number of rockets fired from the Gaza Strip gradually diminished after the ground battle on August 5, when the IDF destroyed 32 offensive tunnels and withdrew ground troops from the Gaza Strip completely (Lappin, 2014). After that, Israel and Hamas engaged in a ceasefire and minor engagements until August 26 (Cohen et al., 2017).

As mentioned at the outset, surveys have been conducted multiple times during the operation and were conducted by the Tami Steinmetz Center for Peace Research, a research institute of Tel Aviv University,⁵ and a joint polling project of the Guttman Center for Public Opinion and Policy Research of the Israel Democracy Institute. Four wartime surveys were conducted on July 14 (hereinafter survey A), from July 16 to 17 (survey B), from August 23 to 24 (survey C), and from August 11 to 12 (survey D). From these surveys, it is possible to compare the evaluation of operational behaviour by air and artillery bombing and ground warfare, respectively. While survey B measured public opinion just before the start of ground warfare, survey C measured public opinion six or seven days before the start of ground warfare. Survey A includes only a sample of Jewish citizens, but surveys B, C, and D also include a sample of Arab citizens, making it possible to analyse political attitudes by ethnicity.⁶

Thus, the dataset allows us to examine the relationship between the rally effect and social identity, as well as changes in the rally around the flag effect as the intensity of war increases. It is well known that exposure to information about the Israeli-Palestinian conflict causes Jews to form aggressive attitudes towards their opponents by viewing Palestinians and Palestinians as ‘out-groups attacking their own communities’ (Feinstein, 2018; Gvirsman et al., 2016). Subsequently, what attitudes do Arabs with Israeli citizenship – Arab Israelis – form? Does the feature of the attitudes of Arab Israelis become clear when Jewish Israelis are taken as the reference group?

Interrupted time-series design

The multiple polls conducted repeatedly makes it possible to analyse the rally effect associated with the entry into the ground war after 17 July 2014, through an Interrupted Time-Series analysis (ITS), a quasi-experimental design that captures the impact of an intervention by taking the passage of time on the horizontal axis and observing the data before and after the intervention at a given time. In the context of Operation Protective Edge, the ground war is an intervention, and the data to be observed is the Israeli government’s assessment. As the Peace Index project measures public opinion in survey B (conducted on July 16–17) just before the start of the ground war and survey C (August 23–24) was a week after the start, a comparison of the two surveys captures the impact of the ground war on the government’s assessment.

The overall trend must be considered to demonstrate that this study is not just a pre-test-post-test design. The government evaluations in survey A, survey B, and survey C were to determine the changes in outcomes over time, and these are illustrated in Figure 3. The horizontal axis is when the telephone survey was attempted, and the vertical axis represents the government evaluation. The left side of the trisection is the outcome distribution of survey A, the centre is survey B, and the right side is survey C. In addition, in Figure 3, the left-hand (survey A) distributed on July 14 (mean 7.8, standard deviation 2.15, n = 244) is similar to the right-hand (survey C) distributed from July 23–24 (mean 8.01, standard deviation 2.02, n = 210 [Jewish sample only]) after the intervention. In other words, the relatively high ratings of the 14 days on the left side of the sample declined (mean 6.36, standard deviation 2.58, n = 181 [Jewish sample only]) during the 16–17-day period in the middle. Therefore, considering the counterfactual that the ground war did not break out and airstrike and artillery fire continued on July 23–24, it can be assumed from the trends in surveys A and B that the government’s assessment would have stayed at the level of July 16–17-day period, or else it would have fallen.

Therefore, this study considers the outcome of survey B conducted on July 16–17 as a counterfactual case in which no ground war was fought and adopts a design that compares this to the outcome of survey C after the start of the ground war and evaluates its impact.⁷ This research design will allow us to adjust for the effects of the covariates and elucidate the relationship between social identity elicited by the ground war rush and the attitudes developed from the rally round the flag effect.

Control variables and multiple assignments

The econometric analysis model in this study is a simple ordinary least squares method. Excluding the independent variables for outcomes – Jewish/Arab differences⁸ and variables representing before and after the entry of the ground war⁹ – variables that could be confounding are gender, age, income, education level, and ideology. Socioeconomic variables ranging from gender to education level represent social stratification in macroscopic terms. It is shared knowledge in quantitative social science that the distribution of political opinions and attitudes is skewed by social strata, and Israel is no exception. The left-right ideology or ideological left-right axis is a concept that is frequently used in everyday and academic discussions of Israeli politics. This policy positioning is well-known to the population because it is useful for political actors such as politicians, parties, and voters to indicate their policy standing in the political discourse space (Caverley, 2014, p. 222; Hamanaka, 2019b).

In addition, when we performed the ordinary least squares method on the original data that combined surveys B and C in the preliminary analysis, we found a significant decrease in the sample. If listwise deletion is used on the missing values in the dataset, we cannot rule out the possibility that they have some effect on the relationship between the outcome and the independent variables. Recent quantitative studies have required us to be concerned with imputation methods of missing observations and the biases they introduce (Takai et al., 2016). Therefore, in this study, to ensure the robustness of the analysis, the missing values are imputed using multiple imputations. The applied method is the traditional algorithm, data augmentation.

Results

The Hypotheses 1a to 3b derived in Section 2 were statistically tested, and the results are presented in Table 2 and Figure 4. As the coefficient for the term distinguishing Jewish/Arab (Jews-Arabs) in (1) of Table 2 is negative, this result negates H1a and supports H1b (Model-1). The value of the constant term indicates that the average government rating by 18-year-old Jewish Israelis is 6.68 points, which is favourable. For 18-year-old Arab-Israelis, the government’s rating for carrying out Operation Protective Edge is 2.14 points, which is negative. To understand this intuitively, the upper left panel of Figure 4 illustrates the relationship between age and government ratings when controlling for the other conditions. The control variables are statistically significant for all socioeconomic variables – gender, age, income, and education level – and even after controlling for these, we can see that the differences between the Jewish and the Arabs have a significant impact on the outcomes.

Table 2. Results of OLS estimation

	(1)	(2)	(3)	(4)	(5)
	Model-1	Model-2a	Model-2b	Model-3a	Model-3b
	Government	Jews	Arabs	Government	Government
Jews-Arabs	−4.5426***	−2.4810***	1.4257***
	(0.4198)	(0.3452)	(0.4362)
Ground Operation				1.4055***	0.2243
				(0.2419)	(0.7527)
Gender	0.5314**	0.1280	0.1137	0.5446**	0.0281
	(0.2307)	(0.1943)	(0.2496)	(0.2313)	(0.7897)
Ideology (Right-Left)	−0.1817	−0.7725***	0.5404***	−0.0891	−0.8258**
	(0.1153)	(0.0987)	(0.1268)	(0.1149)	(0.3979)
Age	0.0247***	0.0229***	0.0049	0.0208***	0.0106
	(0.0064)	(0.0053)	(0.0067)	(0.0063)	(0.0267)
Income	0.2509***	0.1470*	0.1855*	0.0544	0.5250
	(0.0930)	(0.0781)	(0.0982)	(0.0963)	(0.3128)
Education Level	−0.2545***	−0.1923***	−0.0508	−0.2440***	−0.1591
	(0.0878)	(0.0733)	(0.0919)	(0.0908)	(0.2423)
Constant	6.6783***	9.5435***	1.4608**	6.3809***	4.6179*
	(0.5595)	(0.4636)	(0.6045)	(0.5541)	(2.3879)
Observations	459	459	459	401	58

Standard-errors-in-parentheses

*** p < 0.01, ** p < 0.05, * p < 0.1

In Table 2, (2) and (3) are the results of the analyses of Model-2a and Model-2b, which were set up to test H2a and H2b. The outcomes are Jewish and Arab ratings, respectively, and the coefficients of the Jews-Arabs term are negative in Model-2a and positive in Model-2b, respectively, indicating that both H2a and H2b are supported. An intuitive way to understand this is to focus on the Jewish/Arab differences between the changes in the Jewish ratings illustrated in the upper right panel of Figure 4 and the Arab ratings illustrated in the lower-left panel. The control variables were ideology, age, and education level in Model-2a, and ideology in Model-2b, respectively, which are statistically significant.

Columns (4) and (5) of Table 2 illustrate the results of testing H3a and H3b, that is, the analyses of Model-3a and Model-3b. In (4), we ran an ordinary least squares estimation containing only the sample of Jewish Israelis and (5) that of Arab Israelis; the Ground Operation is a dummy variable that represents the pre- and post-ground war entry and is statistically significant only in (4). In other words, H3a, ‘Jewish Israelis raise their assessment of the Israeli government from the decision to execute ground operations,’ was statistically supported, while H3b, ‘Arab Israelis lower their assessment of the Israeli government from the decision to execute ground operations,’ was not. The results presented in (4) of Table 2 are illustrated in the lower right panel of Figure 4, from which we can read that the ground war raised Jewish Israelis’ assessment of the government by 1.4 points, or the ground operation sent citizens into a frenzy.

However, as the Ground Operation term is not significant in (5) of Table 2, it means that Arab Israelis’ evaluation of the government remained low, to begin with, and did not change when they entered the ground war. However, the possibility remains that the sample size of Arab Israelis was small, n = 58, and therefore insufficient to make statistical inferences.

Discussion

Let us return to the question at the beginning of this article: What attitudes do Arab Israelis, a minority in Israeli society, form towards the wartime government of Israel and their own community? According to social psychologists attempting to elucidate the causal mechanism of the rally effect, social identity theory forms the basis of the rally around the flag phenomenon. To highlight the political attitudes of Arabs as a minority, an analysis of the attitudes of Jewish Israelis as a majority was also conducted.

Three hypotheses were derived from social identity theory, and the ordinary least squares estimation by the interrupted time series design yielded the following results. First, Arab Israelis, unlike Jewish Israelis, evaluate governments carrying out war lowly. Next, Arab and Jewish Israelis tend to rate citizens of the same ethnicity highly and citizens of different ethnicities lowly. Finally, only Jewish Israelis produce the rally phenomenon after the start of the ground war, and Arab Israeli evaluations of the government were not affected by the start of the ground war.

Possible confounding factors are addressed by the interruptive time-series design in the analysis. The author believes that the same research design also addresses bias due to missing variables. That said, our study does not tackle the probable process by which Arab media consumption and the claims of Palestinian opinion leaders led to the ‘inverse rally.’ The failure to examine this process may be a limitation of this study. Therefore, a different approach to media information path other than SIT, such as that found in Groeling and Baum (Groeling & Baum, 2008), is a topic for future research.

The empirical analysis above demonstrates that minorities, as ‘hostile subgroups,’ form critical attitudes towards the government and the majority community in a situation when the majority’s assessment of the government follows the rally round the flag effect. As Arab Israelis are likely to identify with the Palestinian community inhabiting the Gaza Strip rather than the state to which they belong, they will probably form a political attitude that takes the side of the ‘enemy’ that will be attacked by the Israel state in times of war. This reinforces the division of Israeli society. Arab Israelis make up 21% of the population (Galnoor & Blander, 2018, p. 632), and even though they are a minority, they cannot be ignored as part of Israeli society.

This study only examined one localised, short-lived war: Operation Protective Edge. Therefore, our findings may not be appropriate in a prolonged war that lasts months or years, and in a large-scale war involving several actors. From the viewpoint of understanding Operation Protective Edge as a phase of the protracted Israeli-Palestinian conflict, it is also a case of collective consciousness being caused by armed conflict. Hence, the causal effects obtained by the analysis in this study might be acknowledged widely for ‘hostile subgroups’ in societies with multiple ethnicities in a state of prolonged conflict. National security emergencies such as war do not involve minority groups by rallying the majority ethnicity under the national flag. Wartime zealotry can repel minorities and exacerbate dividing fissures in society. Thus, the phenomenon of rallying around the flag is similar to a bone stuck in the throat of liberal democracies that contain an ‘enemy subgroup.’

Figure 1. Support rates of PM Netanyahu (%)

Source:Polls conducted by Israeli Media.

Figure 2. Achievements and damages of operation protective edge

Source: Haaretz, 29 July 2014.

Figure 3. Distribution of approval of the government and timing of polls

Figure 4. Estimation results of OLS model

Acknowledgments

This article was originally presented at the 2019 meetings of the Japan Association of International Relations held in Niigata city. The author wishes to thank Yuichi Kubota, Keigo Ohmura, Kazuhiro Obayashi, Gaku Ito, Akira Sato, and Shoko Kohama.

Disclosure statement

No potential conflict of interest was reported by the author.

Additional information

Funding

This work was supported by Japan Society for the Promotion of Science (JSPS KAKENHI) [Grants-in-Aid for Scientific Research] [grant number 19KK0033].

Notes on contributors

Shingo Hamanaka

Shingo Hamanaka is a Professor of Middle East Politics at Ryukoku University in Kyoto, Japan. His research focuses on the strength of authoritarian regimes in the Middle East, public opinion and foreign policy issues in Israel, and Muslim migration from Arab countries. His publications include ‘Demographic Change and its Social and Political Implications in the Middle East’ in the Asian Journal of Comparative Politics (2017), ‘Sensitivity to Casualties in the Battlefield: The Case of Israel’ in the Asian Journal of Comparative Politics (2018), and ‘The role of digital media in the 2011 Egyptian revolution’ in Democratization (2020).

Notes

1. SIT is an important theoretical approach of how the rally effect works. An alternative is the approach focused on the interaction between opposition elites and mass media. Broadcasting positive comments by the opposition party members tend to increase presidential approval rates and negative comments tend to decrease the approval rates (Groeling & Baum, 2008). However, my approach depends on the theoretical base of SIT, not of the interaction between the opposition and media.

2. Hamanaka (2018, pp. 53–55) shows the correlation between support for coalition government and party identification, as a control variable, even in the time of crisis.

3. Hamanaka discusses the struggle between civic and ethnic identities in a slightly different context than citizenship, using the case of Israeli society.

4. If this report is true, it would mean that the 2014 war was a conflict underpinned by the Diversionary Theory by the Hamas, but unfortunately there is currently no way to verify this.

5. The Peace Index site (http://www.peaceindex.org/DefaultEng.aspx) on Tel Aviv University’s server is not available for viewing (as of 14 April 2020). The link is on the Israeli Institute for Democracy’s server (https://en.idi.org.il/centers/1159/1520) and the data can be downloaded from Data Israel (https://dataisrael.idi.org.il/).

6. In this study, which focuses on the effects of rallies during ground combat, only data from Surveys A to C are used, partly because Survey D does not carry over some of the questions common to Surveys A to C, making it difficult to use it for comparative analysis.

7. This is a conservative or under-biased approach to evaluating the intervention.

8. A dummy variable (Jews-Arabs) that distinguishes between Jews (=0) and Arabs (=1).

9. A dummy variable (Ground Operation) that distinguishes between pre entry (=0) and post entry (=1) ground operations.