Presidential popularity and international crises: an assessment of the rally-‘round-the-flag effect in Russia

ABSTRACT

In times of severe international crises, the domestic public typically becomes more loyal to its government, creating a so-called “rally-'round-the-flag” effect and boosting a political leader’s popularity. The bulk of research on the link between international crises and presidential popularity rates deals mostly with the experience of the US and other Western democracies. We focus on the case of Russia as a non-democratic major power. Our empirical analysis investigates whether military conflicts with Russian involvement are followed by an increase in presidential popularity during the presidencies of Vladimir Putin and Dmitry Medvedev. The results of our monthly time-series estimation reveal that a boost in popularity, if there is one at all, is on average small in magnitude for less violent conflicts, whereas a substantially important increase in presidential approval ratings can be observed after international crises in which violence is a central element of conflict management from the very beginning.

KEYWORDS:

• Russian foreign policy
• rally effect
• presidential popularity
• international crises
• military disputes

Introduction

Why countries get militarily involved abroad or escalate international crises can have a number of reasons. Some explanations focus on facets of international interactions between states, such as the distribution and projection of power, or economic interests and interdependencies. Other approaches direct their attention to internal aspects of the participating countries and the influence of relevant domestic actors and institutions, ranging from the interests of economic actors to institutional political restraints. One explanation that repeatedly emerges concentrates on a political leadership’s interest in using international crises to distract from domestic problems with the expectation that military interventions abroad lead to an increase in the political leader’s standing at home. The “wag-the-dog” narrative is prominent in popular discourse and media, but also in academic studies. Several studies linked, for example, the wars in Georgia and Ukraine to Russian diversionary behavior (e.g. Filippov 2009; Gerstel 2016; Vlasenko 2015). Hutcheson and Petersson (2016, 1113) consider the annexation of Crimea a “game changer” insofar as the assertive foreign policy acted as compensation for declining domestic support and less successful economic performance.

The underlying idea is that in times of international crises the general public is unified and rallies behind its political leadership. Whether and to what extent the general public takes interest in matters of foreign policy is fiercely debated. There is strong consensus that public evaluation of a president’s job performance hinges strongly on the state of the economy. However, the state needs to provide prosperity and peace, or as Nickelsburg and Norpoth (2000, 326) conclude from their study “the Commander-in-Chief counts as much for public approval as does the Chief Economist.” The expectation that the president’s popularity will rise after very “specific, dramatic, and sharply focused” international events in which the president is directly involved (Mueller 1970, 21) is frequently referred to as the “rally-'round-the flag” effect. While there are numerous examples of a rally effect (Nickelsburg and Norpoth 2000), a knee-jerk public reaction and burst in patriotism in light of an external threat is far from certain. Several studies only found limited support for such an anticipated boost in popularity (e.g. Baker and Oneal 2001; Lai and Reiter 2005), thus raising doubts that it would serve as a good foundation for a political leader’s presumed diversionary behavior (Murray 2018). A rally effect is typically observed after severe conflicts and seems to be a short-term phenomenon in most cases. On the flip side, political leaders involved in a costly war are also confronted with a public that is averse to casualties, which bears the risk of declining public support and the loss of elections in the mid or long term.

Public support for the government is not only important in a democracy; the leadership’s high popularity also serves as a source of legitimacy in autocratic political systems. Public opinion does play a role in Russian politics and the Kremlin invests heavily in persuading ordinary Russians of its policy and in suppressing dissent. Russia was widely considered an electoral authoritarian system for many years, and Russian presidents relied on popular support as a major source of legitimacy. In recent years, the regime, however, has become increasingly repressive towards the opposition and independent media, also in light of economic difficulties, declining popular support, and rising protests (Petersson 2021). Still, even political regimes that rule with an iron fist need public support to some extent, or at least have an interest in avoiding widespread public opposition. The popularity of Vladimir Putin over the years has been closely watched inside and outside of Russia, especially surrounding evolving international crises.

Russia is a particularly interesting case to study the rally effect after military conflicts because as a major power it has the capacity to deploy its troops abroad and has done so on numerous occasions in the past. The collapse of the Soviet Union had generated big potential for conflicts in the post-Soviet space and triggered the process of a redistribution of power in the region. In recent decades many conflicts have flared up between Russia and its former fellow Soviet states, which wanted to gain independence from Russia’s influence and were striving for more integration with Western countries. Most international conflicts in recent Russian history and discourse have been placed in the framework of tense relations between Russia and Western countries. So that in a sense “the West” can be considered as the most significant “other” for Russia in relation to whom the country’s foreign policy ambitions have been constructed (Tsygankov 2012), thus conforming to the in-group/out-group dynamics that are a central element of the sociological explanation of a rally effect, stressing the argument that an external threat strengthens internal cohesion.

Whether the Russian president’s popularity actually benefits or suffers from involvement in international crises is an open question. In this paper, we investigate more systematically if engagement in international crises has an impact on presidential approval ratings in Russia with a quantitative analysis of monthly time-series data. Our results contribute to the general debate on the relationship between military interventions and public opinion and will provide an empirically based assessment of the question of whether political leaders are well advised to count on a boost in popularity as a result of a military operation. Our findings for Russia correspond to findings on other countries, which suggest that only a few military conflicts trigger a rally effect that is sufficiently strong and sustainable (Baker and Oneal 2001; Feinstein 2022; Mueller 1970; Murray 2018). Most conflict events are not followed at all, or only by a small rise, in popularity. A substantially important rally effect can only be observed after international crises with violence as a central element to conflict management already at the beginning of the crisis.

Presidential popularity after international crises

While the public in the long run is generally opposed to conflicts that bear high casualties (Baum and Potter 2008; Gartner and Segura 2000), there is at the same time the notion that in the short run, in light of external threats and severe international crises, the public will set aside possible criticism and disagreements with the government and throw their support behind their soldiers and the commander-in-chief (Baker and Oneal 2001; Mueller 1970).

Not all military operations have the potential to capture public attention. Routine events often go unreported, and not all non-routine events are important enough to the public either to trigger a rally. Mueller (1970) argues that only clearly focused events that occur suddenly are suited to get the public’s attention, whereas gradual developments fail to do so. While early case studies on public support for the use of force have been criticized for selection bias (e.g. Jentleson 1992), subsequent studies were more systematic about testing the rally effect by investigating all international crises, or more restrictively, all major uses of force in foreign policy crises (Oneal, Lian, and Joyner 1996). The focus on major crises avoids the pitfall of including minor military incidents that hardly are known to the public. Military operations need to be salient to the general population to trigger a rally effect, especially if the public is considered to be “pretty prudent” and critical with regard to the policy objectives that are pursued (Jentleson 1992; Oneal, Lian, and Joyner 1996). In several empirical studies, a sizable increase in popularity is observed only in case of very severe crises or conflicts that eventually escalate to full-scale war and are a threat to the national interest (Baker and Oneal 2001; Lai and Reiter 2005).

The rally effect is a short-term phenomenon and applies mostly to immediate reactions of the public to the initiation of the use of force. It is thus important to focus on the outbreak of an international crisis or major events that escalate an ongoing and protracted conflict to a higher level. At the outset of a foreign policy crisis, the public is at an information disadvantage and therefore takes its clues mostly from the actions and reactions of the political leadership, especially as long as the media and opposition do not provide different insights. Typically, in case of a fast-unfolding international crisis, the opposition sets aside its criticism in order to not appear unpatriotic. Once more balanced information becomes available, the surge in public approval ceases (Baker and Oneal 2001).

Previous studies that analyzed the popularity of Russian presidents focused on economic factors, domestic order, and terrorist events (Fedotenkov 2020; Hutcheson and Petersson 2016; Treisman 2011). We are interested in finding out whether Russian presidential popularity increases in light of external threats and the outbreak of an international military crisis more generally, and in particular whether only highly salient and severe conflicts with violence as a central component of conflict management have the potential to rally the public behind the political leadership. Based on previous empirical studies of other country cases, we expect that the rally effect is not automatic after all international military crises but substantially significant for the outbreak of more violent and severe crises only. Our focus will be on the beginning of international crises that are triggered and/or responded to with elements of violent conflict management and on conflicts that escalated to an extent that the conflict parties crossed the threshold of using violence.

Research design

We will provide a descriptive assessment of the relationship between Russian presidents’ popularity and international military conflicts, followed by a quantitative multivariate time-series analysis. The focus of the quantitative analysis will be on the rally effect, and thus an examination of the claim that the outbreak of an international conflict with Russian involvement will be related to an increase in a president’s popularity in the month after. We will draw on time-series data on a monthly basis covering the years 2000–2022. In order to limit the scope of the investigation, only conflicts that started in the time period during the presidencies of Vladimir Putin and Dmitry Medvedev will be taken into account.¹ For most tests, we take the year 2020 as the cut-off point, because of concerns regarding data quality for the more recent years.

Until 2020, Russia was mostly classified as a hybrid regime or electoral autocracy (e.g. Dollbaum 2020; Lankina and Tertytchnaya 2020; Rosenfeld 2018). The adoption of the constitutional amendments strengthened presidential authority and allows Putin to stay in power for years to come, thus cementing the political regime’s transition from an electoral autocracy to a personalist dictatorship by 2020. The following year was additionally experiencing a vast crackdown on independent and free media and internet outlets (McFaul 2021). With its transition to a personalist dictatorship without institutional possibilities to restrain the leader (Gomza 2022), as other researchers, we consider the representativeness of public opinion polls under dictatorial regimes as problematic. According to Kizilova and Norris (2022), there are many factors that can lead to response bias under dictatorial regimes, including state and self-censorship, the emigration of dissenters, massive repressions, manipulation of public opinion, and propaganda. Therefore, such public opinion polls should be treated with great caution.

For our variables of main interest, international crises with Russian involvement, we rely on two data sources: the Heidelberg Institute for International Conflict Research’s (HIIK) Conflict Barometer (2023, 9–10) and the International Crisis Behavior (ICB) Data (Brecher et al. 2021). The HIIK data are available until 2022 and define political conflicts as incompatible actions and acts of communication concerning conflict issues, such as resources, territory, or ideology among others, that are relevant to society and that threaten the functioning of a state or the international order (HIIK 2023). Distinguishing among five levels of intensity, the HIIK data include violent and non-violent events and are thus one of the most comprehensive data sets on international conflicts. We focus on conflicts that involve violence, defined as the deliberate exertion of physical injury. The variables will take the value of one if we observe the start or escalation to an international violent conflict of medium or high intensity in a given month and zero otherwise. We will thus differentiate between HIIK violent conflicts of medium intensity and HIIK wars of high intensity, as described in the discussion below on data sources.²

In additional tests, we draw on the widely used ICB data (Brecher et al. 2021), which consider a foreign policy crisis as triggered by a “specific act, event or situational change which leads decision-makers to perceive a threat to basic values, time pressure for response and heightened probability of involvement in military hostilities” (Brecher and Wilkenfeld 1997, 5). The data provide different categories of triggering events, ranging from verbal and economic acts to military acts. Our variable for crisis trigger takes the value of one in a month when a crisis was triggered and zero otherwise. The ICB data also provide information on the entity that triggered the crisis as perceived by the conflict actor. In all our cases, Russia as conflict actor perceived another actor as responsible for triggering the crisis and responded in most cases within the same day or a few days later. This response can take, similar to the trigger variable, different values, such as inaction, a verbal response, or conflict management responses to a crisis that are coded as non-violent or violent military acts. From the data, we generate dummy variables for violent and non-violent crisis response that take into account whether the use of violence was central to the decision-maker in his/her crisis management technique, with the use of minor violence as the reference category.

While the Heidelberg Conflict Barometer allows us to cover the years 2000–2022, the ICB data are currently available only until 2019. For the Russian Federation during our period of study, the HIIK data include 10 conflicts and the ICB data contain eight conflicts (see Appendix Table A1). The conflicts in both data collections partly overlap: both include the three cases between Russia and Georgia in the years 2002 (Pankisi Valley), 2004 (South Ossetia), and 2008 (Russian-Georgian war), the annexation of Crimea and the conflict in Donbas in February 2014, the Turkey–Russia jet incident in 2015, and the Kerch Strait dispute in the Black Sea in November 2018. Both data collections provide information on Russian involvement in the Syrian war; in the case of the HIIK data, this is the beginning of Russian air strikes in support of the Syrian government in late September 2015, whereas the ICB data include the Syrian chemical weapons crises in 2017 and 2018. The HIIK data, furthermore, include border disputes with Japan in 2006 and Kazakhstan in 2013, and given its longer time range, includes the February 2022 invasion in Ukraine. With the exception of the war with Georgia in 2008, which took place under the presidency of Dmitry Medvedev, all other conflicts occurred during the presidential terms of Vladimir Putin.

The question of whether the general public is supportive of or adverse to military interventions has been investigated for individual military crises and mostly relies on survey data that are often not easily comparable. Comparative studies over time, on the other hand, provide the political elite with information about trends and changes in support before, during, and after military crises and not just as a one-time snapshot. However, such studies face the challenge of having to rely on information from surveys that are regularly repeated with identical questions and not tailored toward a specific conflict. A common approach is to draw on the popularity ratings of country leaders. In some countries, presidential public approval ratings are a continuing assessment of how satisfied the population is with its leadership and their job performance.

For the Russian president’s approval rating, we rely on polls conducted by the renowned Levada Center, a Russian independent non-governmental research organization that is widely used in international scientific research (e.g. Treisman 2011). It has been thoroughly investigated with list experiments in 2015, providing support for the respondents’ genuine answers without severe social desirability bias (Frye et al. 2016), a conclusion that warrants, however, more caution in the replication of the study in late 2020 and early 2021 (Frye et al. 2023). Concerns about bias in the answers for fear of repression are valid, but we still can expect, starting from maybe overall too high levels, to get information about trends and changes (Petersson 2021). The question addressed by Levada is formulated as the following: “Do you approve the activities of Vladimir Putin as the President of Russia?” (Levada Center 2022).³ In robustness tests, we also utilize data from the Public Opinion Foundation (FOM 2023a, 2023b), a polling agency closer to the state that regularly asks, “Do you think that President Vladimir Putin is doing his job rather well or rather bad?”

Public opinion polls are a convenient measure to assess a government’s popularity and express whether the general population is more or less satisfied with how the government is doing its job. However, low public support is just the expression of more underlying factors. While our focus is on international conflicts, presidential popularity rates will vary with domestic conditions. Previous research has shown that especially economic conditions, or their perception, matter for the popularity of Russian presidents (Fedotenkov 2020; Treisman 2011) or the popularity of the ruling party (Rosenfeld 2018). Countries with low growth rates, high unemployment, or inflation often observe a higher risk of regular, and even irregular change in government. In severe economic situations, the incentive to engage in diversionary behavior is particularly high, and consequently many studies focus on economic difficulties (Brulé and Hwang 2010; DeRouen 1995; Oneal and Tir 2006). Therefore, we will account for the monthly growth rate in the consumer price index relative to the previous month to capture price increases, the normalized and seasonally adjusted monthly GDP growth rates from the previous month, and the seasonally adjusted registered monthly unemployment rate in our multivariate analyses as control variables. For the economic indicators, we use data from the OECD (2022a, 2022b, 2022c).

In addition, we expect a president’s popularity to suffer in light of mass demonstrations and protests, which are another form of how the population can voice its discontent outside the election cycles. We use Mass Mobilization Protest Data, which include protests of at least 50 people that are targeted against the government or state policy for all countries in the time period 1990–2020 (Clark and Regan 2016). Together with economic stagnation, mass protests could trigger political instability and reflect negatively on the government’s popularity, potentially providing a reason for leaders to divert public attention to international issues (Kisangani and Pickering 2011). The electoral cycle also matters. Following others’ research on presidential popularity (Anderson 1995; Lai and Reiter 2005), we control for a potential honeymoon effect for newly elected presidents and the number of months an incumbent has been in office, with the expectation of a decline in support over time. Honeymoon captures the higher but declining popularity newly elected presidents enjoy in the first months after the election. In the 6 months after an election, it is coded from +6 down to +1 and zero otherwise. Months in office is a count variable for every year the president holds office.

A descriptive assessment of rally effects after international crises

Before we move to the multivariate analysis, we present some descriptive accounts of the international conflicts and their potential rally effects. The line in Figure 1 marks the monthly presidential approval ratings and the dots indicate the months when a specific violent conflict took place in case of very brief crises or started in case of longer conflicts in a given month. Overall, we observe what others have emphasized before: Vladimir Putin enjoys generally high popularity rates (Frye et al. 2016; Hutcheson and Petersson 2016; Treisman 2011).

Figure 1. Monthly presidential approval ratings, 2000–2022.

The graph reveals that not all but only some crises and conflicts are followed by a rise in popularity rates that indicate a clear rally effect. In particular for the conflicts at high levels of violence and where major violent acts were central in the first days, namely Ukraine in 2014 and 2022 and Georgia in 2008, we can observe significant rally effects of more than 10 percentage points. Other crises involving Georgia or the intervention in the Syrian civil war were followed only by a small increase in approval rates, and for several minor crises no change or even a decline in popularity can be observed when Russia responded to the crisis with non-violent means only. Large rally effects

Figure 1 reveals a very significant and sustainable rally effect after the annexation of Crimea in February 2014 and after the Russian invasion in Ukraine in February 2022. Both events were followed by very similar patterns in Putin’s approval ratings, starting from: (1) some of the lowest levels of his popularity; (2) an important jump in approval ratings after the conflict started or escalated to violent levels, and (3) the sustaining of high rates for an extended time period. The Maidan protesters ousting the government on 22 February 2014 is considered the trigger of the crisis from a Russian perspective. Responding with a violent military act within days, Russian special forces seized government buildings in Crimea, and on 1 March a full-scale military intervention started, followed by the formal annexation and a referendum by mid-March. Shortly afterwards, the conflict in the Donbas region started (Brecher et al. 2021). A large majority of Russians supported the incorporation of Crimea into Russia. After the annexation of Crimea, Putin’s approval ratings increased by up to 20 percentage points and his popularity stayed continuously high for 4 years. It also had a major impact on Putin’s popular support among those citizens who usually do not consume state-controlled television and drew attention to politics among parts of the population who were not politically involved (Hale 2018, 369).

The attack on Ukraine in February 2022 was followed by a very similar pattern. Tension with Ukraine and NATO was mounting already in the beginning of the year with the Russian military build-up along the border. After the Russian invasion, violence in this protracted conflict escalated to a full-fledged war (HIIK 2023). Putin’s popularity was in the high 60s before the intervention and increased to 83% in March 2022 and remained high beyond the period of study. As discussed above, the quality of public opinion polls, especially since 2020, should be taken with a grain of salt. According to Yudin (2022), before the Ukraine invasion Russians could more easily oppose or ignore the government’s political course; since the war they have been increasingly required to show active support. Fear of stronger repression can contribute to the high public support.

We also observe patterns of a rally effect related to the protracted conflict with Georgia. In particular, the Five-Day War in 2008 was followed by a 10-percentage-point increase in popularity of President Medvedev. After Russia announced an increase in its peacekeeping troops in Abkhazia and an alleged troop movement to South Ossetia, Georgia mobilized its reserve. Thus, the triggering event as well as the major response on 8 August were violent acts (Brecher et al. 2021). However, the rally effect after the Russo-Georgian war was not of long duration and Medvedev’s approval ratings dropped rather fast in the following months back to and even below pre-war levels.

Small rally effects

The earlier crises with Georgia were followed only by smaller rallies, not a very pronounced surge and not of long duration. Although the crisis related to the Pankisi valley in August 2002 and to neighboring South Ossetia in August 2004 was triggered and responded to by violent acts, only minor violence and minor clashes characterized management of the conflict. Russia did not go beyond verbal acts in major response to the crisis trigger (Brecher et al. 2021). Both confrontations with Georgia did not lead to a significant rally effect. Presidential popularity remained stable at first after the beginning of the crisis over the Pankisi valley and in the following months increased in parallel with the verbal escalation. After the tensions in South Ossetia, Putin’s popularity increased slightly by four percentage points but only for two months before it returned to its pre-conflict level.

The Russian intervention in the Syrian civil war is treated differently in both data sets. For the Heidelberg Conflict Barometer, this is an intervention into a full-fledged war and thus a high-level violent conflict. At the end of September 2015, Russia started numerous airstrikes in support of the Syrian government (HIIK 2015). Russian military involvement moved the Syrian civil war into a phase of new violent escalation (Osborn and Stewart 2015). Although President Putin’s decision to send military aircraft to attack allegedly ISIS forces in Syria was supported by 66% of Russians (VCIOM 2015), public support for the intervention did not translate into a substantial increase in approval ratings. Putin’s popularity increased for only a few months and continued to decline thereafter.

No rally effects

There are several conflicts that were not followed by any boost in popularity. The ICB data focus on the crisis between the US, Syria, and Russia related to Syria’s use of chemical weapons in April 2017 and April 2018, and subsequent US air strikes. Russia responded to this violent crisis trigger with verbal assurance of support for Syria and a veto in the UN Security Council, but not with violent means (Brecher et al. 2021). The 2017 crisis is not followed by any change in presidential approval ratings and after April 2018 popularity even takes a deep dip.

Popularity rates did not increase after the incident in November 2015 when Turkey shot down a Russian fighter jet, which was on its way to Syria (HIIK 2015). While the crisis was triggered by a violent act, Russia responded immediately in a number of ways, including the announcement of economic sanctions, canceling a state visit, and deployment of an air defense system in Syria, but did not opt for violent conflict management. Instead, it announced the next day that it would not go to war with Turkey over the incident (Brecher et al. 2021). In spite of high media attention, Putin’s popularity ratings did not change after this incident but continued to decline.

The dispute with Japan over the Kuril Islands in August 2006, when the Russian coast guard killed a Japanese fisherman and arrested three others who had entered the disputed territory (HIIK 2006) was even followed by a decline in approval ratings. After the violent incident, Russian officials expressed their regret over the death of the fisherman, but attributed responsibility to the Japanese authorities (McCurry 2006). Russian media did not report much about this incident, so that the short drop in the president’s popularity of three percentage points in the following months could be related to other factors.

We can hardly speak of a rally effect also in connection with the dispute in November 2013 over a border violation in the Caspian Sea, when Kazakh border guards fired at two Russian boats. As a reaction, Russia condemned the use of firearms and demanded a detailed investigation of this case (HIIK 2013). Popularity rates also remained in the 60s after November 2018, when Russian guards opened fire on Ukrainian military boats passing the Kerch Strait. Both countries accused one another of escalation of the incident. The sea border violation led to the temporary declaration of martial law in several regions of Ukraine but otherwise was followed by an increase in military drills only (HIIK 2018). Given the Russian non-violent response, it comes as little surprise that this crisis did not have the potential to a rally effect. This descriptive exercise indicates that only very few and most notably rather severe and violent conflict events seem to trigger a significant rally effect.

Results of the multivariate analyses

In the next step, we go beyond a simple descriptive assessment and test the effect of international conflicts and military crises with Russian involvement on presidential popularity in a multivariate analysis holding other factors constant. Following other studies of presidential approval ratings (e.g. Treisman 2011) and as indicated by various test statistics,⁴ we estimate autoregressive fractionally integrated moving-average (ARFIMA) models that are appropriate for long-memory processes where the autocorrelation functions decay slowly. The fractional-difference parameter, d, is rather large and statistically significant in Table 1, thus supporting the use of ARFIMA models. Furthermore, the ar(1) parameter is also statistically significant, indicating some short-range dependence, whereas the moving-average component did not add explanatory power.

Table 1. Presidential approval ratings (Levada).

	HIIK data				ICB data
	2000–2022		2000–2020		2000–2019
Model:	(1)	(2)	(3)	(4)	(5)	(6)
Conflictt-1	2.726		2.358
	(0.729)**		(0.759)**
Conflict, medium, t-1		1.227		1.306
		(0.852)		(0.852)
Conflict, high, t-1		6.233		5.984
		(1.300)**		(1.593)**
Crisis trigger, t-1					2.250
					(0.825)**
Crisis, violent response, t-1						6.440
						(1.608)**
Crisis, non-violent response, t-1						0.780
						(1.017)
Months in office	0.022	0.023	0.021	0.021	0.028	0.029
	(0.022)	(0.021)	(0.023)	(0.023)	(0.023)	(0.023)
Honeymoon	−0.156	−0.152	−0.163	−0.158	−0.175	−0.161
	(0.336)	(0.330)	(0.354)	(0.350)	(0.353)	(0.348)
CPI growth			−0.025	−0.070	0.110	−0.006
			(0.407)	(0.402)	(0.413)	(0.405)
Constant	71.146	71.110	70.039	70.040	70.182	70.247
	(2.997)**	(3.005)**	(11.122)**	(10.827)**	(11.428)**	(10.997)**
ar(1)	0.952	0.953	0.451	0.459	0.412	0.428
	(0.023)**	(0.022)**	(0.097)**	(0.101)**	(0.090)**	(0.095)**
d	−0.112	−0.108	0.468	0.466	0.472	0.470
	(0.060)	(0.060)	(0.051)**	(0.055)**	(0.042)**	(0.047)**
Sigma 2	9.037	8.702	9.710	9.457	9.708	9.362
	(0.776)**	(0.747)**	(0.877)**	(0.855)**	(0.896)**	(0.864)**
N	271	271	246	246	236	236

Notes: Coefficient and standard error in parentheses, *p < 0.05; **p < 0.01.

In the first two columns of Table 1 we report the results of the HIIK data for the time series until December 2022. The conflict variable, which is lagged by one month, is positively and significantly related to presidential approval ratings. The coefficient indicates that, on average, presidential approval rates increase by 2.72 points after a month with a conflict. Thus, the results suggest that approval ratings are significantly higher in the month after an international conflict in which Russian involvement occurred. In the model reported in the second column, we distinguish between violent conflicts of medium intensity and wars of high intensity. As expected, the rally effect is clearly stronger after violent conflicts of high intensity. The coefficient indicates that on average the approval rate is 6.23 points higher and the result is statistically highly significant. The variable accounting for violent conflicts of medium intensity only is not significant at conventional levels and the coefficient is only 1.22. In models 3 and 4, we end the analysis in 2020 because of concerns about data quality (discussed above) and because of the availability of more control variables. The results are quite similar to the longer time series, but the effects are slightly less pronounced with smaller coefficients for conflicts and high-intensity conflicts. We resort to the data on International Crisis Behavior in columns 5 and 6, which is only available until 2019. The finding is essentially robust, again with a highly significant and positive coefficient, especially for international crises that incorporate violence as a central conflict management technique already in the first days. In model 5 the variable for a crisis trigger more generally is positive and highly significant. In model 6, we differentiate whether Russia responded to the crisis trigger with violent or non-violent means. As before, the coefficient for events with a more violent response is much larger in magnitude than for less violent crisis responses, 6.44 vs. 0.78 points.

The control variables are mostly insignificant. We do find some support for the expectation that the longer a president is in power the more popular he/she gets. Contrary to studies of other countries, we do not find a honeymoon effect in Russia; the coefficient is negative.⁵ Using the monthly growth in the consumer price index to capture economic discontent also is not significantly related to presidential popularity at conventional levels, but shows (as expected) a negative coefficient in most models. This variable (as several others) was not yet available for the longer time series and thus could not enter the model specification in columns 1 and 2.

We conducted a battery of tests of robustness (see Appendix Table A2) with different model specifications, including additional control variables. We accounted for further aspects in the electoral calendar such as a control for presidential elections and backswing. Instead of a growth in consumer prices, we also tested monthly economic growth and unemployment rates to capture economic problems. Higher economic growth has a positive coefficient and higher unemployment is related to declining popularity, but both variables are statistically not significant. Finally, the number of protests (also if we limit the protests to very large ones of more than 5,000 protesters) is not related to presidential popularity. None of the additional control variables turned out to be statistically significant. More important, their inclusion did not change our variables of main interest. The general finding that more violent conflict is related to higher approval ratings has also been supported in a model specification that controls for the months when Vladimir Putin was president. The dummy variable indicates that Putin is generally more popular than Medvedev but is not statistically significant and its inclusion does not change our main finding. We furthermore tested presidential approval ratings based on data from the Public Opinion Foundation, a pro-state public opinion institute in Russia (FOM 2023a, 2023b); the results were not statistically significant but the size of the coefficients also suggests a larger increase after the onset of more violent conflicts.

Finally, we assess whether conflicts that involve territorial issues lead to stronger rally effects. International conflicts that are associated with territorial issues have higher potential to create strong feelings of national unity. Such conflicts are often more intense and arguably more salient to the population and enhance identity formation along in- and out-groups, strengthening national unity. Territorial conflicts are typically perceived by the population as serious threats to society (Gibler, Hutchison, and Miller 2012). Based on the Heidelberg Conflict Barometer data, we test a dummy variable that accounts for whether the conflict item involves territory, namely a dispute over the demarcation of an international border. This is the case for the annexation of Crimea and the Kerch Strait incident as well as the border conflicts with Japan and Kazakhstan. The dummy variable, however, was not statistically significant, not for the time series until 2020 and not if we tested a longer time frame that included the 2022 Russian invasion in Ukraine.

In sum, we find support for the expectation that presidential popularity increases after the start of international conflicts and crises. However, in most conflicts, the rally effect is of a quite small magnitude. Such a small rally effect can hardly be sufficient motivation for a political leader to engage in a potentially costly and risky conflict for diversionary purposes. Only in very severe conflicts that involve very violent elements of conflict management from the beginning did our results reveal a strong boost in the president’s popularity.

Discussion and conclusion

This study provides an assessment of the relationship between international crises and presidential popularity in Russia, based on the analysis of 10 international crises and violent conflicts that took place during the presidencies of Vladimir Putin and Dmitry Medvedev. Our results support an increase in popularity after international conflicts on average. However, the rally effect is not automatic and cannot be observed in all cases. More important, the boost in popularity, if there is one at all, is on average small in magnitude. A strong boost in popularity can only be observed after the outbreak of very severe international crises and conflicts when Russia responded with violent means. The wars in Georgia in 2008 and in Ukraine in 2022 and the annexation of Crimea in 2014 are the only conflicts in our period of study that provide support for a pronounced rally-‘round-the-flag effect of more than 10 points. We detected a significant increase after the Russian-Georgian war, which ceased to exist after a few months. A strong and enduring rise in presidential popularity emerged after the annexation of Crimea and after the 2022 full-fledged invasion in Ukraine. For other conflicts that involved minor levels of violence or when Russia responded with non-violent conflict management techniques, there is only a small rise or no boost in popularity. Thus, our study for Russia supports studies of other countries that detect a strong rally effect only for severe conflicts with high levels of violence (e.g. Lai and Reiter 2005).

Severe and intensively fought conflicts, however, carry the risk of further escalation and long duration. The loss of casualties is a major concern for political leaders who are eager to win conflicts quickly before public support wanes (Bennett and Stam 1996). Long-lasting conflicts that are costly in terms of casualties and economic hardship risk losing the support of the population that is typically not willing to carry such costs. Surveys during the Second Chechen War underline that the Russian public is also predominantly sensitive to casualties among its own soldiers and was alarmed at the economic costs (Gerber and Mendelson 2008). The war in Georgia and the annexation of Crimea were relatively short operations with relatively few casualties, minimizing the risk that a casualty-averse public would turn against the Kremlin. This is different with the 2022 war in Ukraine, which carries immense human and economic costs for the Russian population. While the short rally effect after the war in Georgia corresponds to patterns found in other countries (e.g. Baker and Oneal 2001), the enduring increase in presidential popularity after the irredentism in Ukraine is striking.

A vital factor regarding rally dynamics is the independence of media, which plays a major role in whether the information gap between the government and the public can be narrowed or whether the public will only be exposed to the official state narrative. A public with little access to independent media is likely to embrace the Kremlin propaganda and the narrative of conflicts that were provoked by Russian adversaries and proclaimed threats to the Russian-speaking community. This framing against the out-group of NATO and Western states helped in rallying the public behind its leadership (Theiler 2018; Vlasenko 2015).

Given the uncertainty of a rally effect, one can indeed be sceptical of whether political leaders provoke a potentially costly conflict mainly for diversionary purposes. Other motivations, such as geopolitical interests and Russia’s interest in maintaining its position in the post-Soviet space and support of Russian-speaking communities outside of Russia, cannot be dismissed. Instead, political leaders might use emerging crisis situations and possibly reframe them with their interest of a boost in popularity in mind.

Data availability statement

The replication data and do-file for the empirical analysis are available at https://doi.org/10.7910/DVN/2I0RW8

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

Work on this article was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – GRK2560, project number 413881800.

Notes

1. For this reason, the Second Chechen War which started under the presidency of Boris Yeltsin, is not part of our analysis.

2. None of the conflicts with Russian involvement during our period of study carries the category of 4 (limited wars) and only 3 conflicts are categorized as full-scale wars, namely the war with Georgia in 2008, the intervention in Syria in September 2015, and the 2022 Ukraine invasion.

3. Levada does not report an approval rate for June 2018 and December 2020. For the arfima analysis, however, we need an uninterrupted time-series and for this reason interpolate this value. The pattern of a continuous decline corresponds to the pattern in the FOM data.

4. We checked for unit root with autocorrelation and partial autocorrelation plots and various tests. The augmented Dickey-Fuller test is with -3.536 above the 1% critical value and thus we cannot reject the null hypothesis that the time-series is non-stationary. Furthermore, based on the Kwiatkowski-Phillips-Schmidt-Shin test of 0.083 that is less extreme than the critical value of 5% we cannot reject the null hypothesis that the timeseries are trend stationary.

5. Months in office and the honeymoon variables are not very highly correlated (r = -.19) and the variance inflation factor test also does not provide reason for concern. The overall mean vif is 1.19 and thus far below the recommended cutoff value of 10 or even 5. The individual vif for months in office is 1.37 and for honeymoon 1.11.

Appendix

Table A1. List of international violent conflicts with Russian involvement (HIIK and ICB data).

Violent conflict	Month	HIIK	ICB
Georgia (Pankisi valley)	July/Aug 2002	x	x
Georgia (South Ossetia)	Aug 2004	x	x
Japan (Kuril Islands)	Aug 2006	x
Georgia (South Ossetia)	Aug 2008	x	x
Kazakhstan et al. (Caspian Sea)	Nov 2013	x
Ukraine (Crimea and Donbass)	Feb 2014	x	x
Syria intervention	Sep 2015	x
Turkey	Nov 2015	x	x
Syria (chemical weapons)	April 2017		x
Syria (chemical weapons)	April 2018		x
Ukraine (Kerch Strait incident)	Sept/Nov 2018	x	x
Ukraine	Feb 2022	x

Table A2. Presidential approval ratings, 2000–2020, additional tests.

	Presidential approval rates (Levada)						(FOM)
Conflict, med,	1.296	1.310	1.298	1.307	1.299		0.655
t-1	(0.854)	(0.852)	(0.863)	(0.852)	(0.853)		(0.999)
Conflict, high,	5.967	5.993	5.971	5.985	5.997		1.265
t-1	(1.596)**	(1.594)**	(1.616)**	(1.592)**	(1.594)**		(1.871)
Months in	0.022	0.021	0.026	0.022	−0.014	0.021	0.025
office	(0.023)	(0.023)	(0.023)	(0.023)	(0.041)	(0.024)	(0.031)
Honeymoon	−0.177	−0.165	−0.158	−0.164	−0.294	−0.179	0.244
	(0.351)	(0.350)	(0.353)	(0.351)	(0.371)	(0.360)	(0.429)
CPI growth			0.010	−0.072	−0.095	−0.040	0.287
			(0.408)	(0.402)	(0.402)	(0.416)	(0.478)
Unemployment	−0.696
	(1.198)
GDP growth		120.464
		(144.336)
Protest			−0.034
			(0.166)
Backswing				−0.114
				(0.852)
Presidential				0.773
elections				(2.015)
Putin					3.796
					(3.594)
Territorial						1.249
conflict, t-1						(1.178)
Constant	71.297	70.096	70.113	70.042	70.118	70.130	51.545
	(11.049)**	(11.040)**	(11.084)**	(10.815)**	(10.646)**	(12.017)**	(3.640)**
ar(1)	0.454	0.454	0.435	0.460	0.454	0.427	0.917
	(0.100)**	(0.098)**	(0.097)**	(0.102)**	(0.101)**	(0.087)**	(0.045)**
d	0.467	0.468	0.470	0.466	0.466	0.473	−0.007
	(0.054)**	(0.052)**	(0.049)**	(0.056)**	(0.055)**	(0.042)**	(0.093)
sigma2	9.446	9.430	9.486	9.448	9.416	10.034	12.727
	(0.854)**	(0.853)**	(0.872)**	(0.854)**	(0.851)**	(0.907)**	(1.222)**
N	246	246	238	246	246	246	217

Note: Coefficient and standard error in parentheses, *p < 0.05; **p < 0.01.