No Zeitenwende (yet): Early Assessment of German Public Opinion Toward Foreign and Defense Policy After Russia’s Invasion of Ukraine

This paper addresses the question of whether Russia’s invasion of Ukraine led to a turning point (Zeitenwende) in public opinion on foreign and defense policy in Germany. To this end, we provide a theoretical analysis of how the concept of turning point can be applied to public opinion. We identify the durability of the change in attitudes as well as its significance as necessary conditions to speak of a turning point. In the remainder of the paper, we focus on the argument that changes in different types of orientations are significant to different degrees. Change in core postures is more significant than change in policy attitudes; change in attitudes thematically distant from the Russian invasion is more significant than change in attitudes directly related to the event. Empirically, we present a panel data analysis of attitude change triggered by the Russian invasion. Analysis of data from several waves of the German Longitudinal Election Study (GLES) panel survey collected before the invasion (2017–2021) and in two waves after (May and October 2022) shows that there were sizable shifts in policy attitudes directly related to the event. Postures remained essentially unchanged, as did thematically distant attitudes. We conclude that there has been no turning point at the level of public opinion (yet).


Introduction
The claim that Russia's invasion of Ukraine in 2022 was a Zeitenwende quickly became an anchor of public debates in Germany. Chancellor Scholz popularized the term when he used it just days after the invasion began. "[Zeitenwende] means," Scholz said, "that the world will not be the same afterwards as it was before" (Scholz 2022). Accordingly, the invasion represented a fundamental break with the (liberal) international order that required Germany to revise some of its basic foreign policy beliefs and policies. Scholz elaborated on this by announcing, among other things, arms deliveries to Ukraine, a significant increase in military spending, and the strengthening of cooperation with like-minded countries. The following debates focused on what effects the Russian invasion had or should have on the international order and German foreign policy, and whether these are correctly described by the term Zeitenwende (e.g., Blumenau 2022;Bunde 2022;Groitl 2022;Fröhlich 2023;Risse 2022).
Less systematic attention has so far been paid to the question of how the German population reacted to the Russian invasion. Was there a Zeitenwende at the level of German public opinion? This question is highly relevant, since democracies are characterized by a complex interrelationship between citizens and political decisionmakers. Changes of course initiated from the top not only gain democratic legitimacy, but they are also more likely to last if they are broadly in line with public opinion. If they are not, and decision-makers fail to persuade the public of their position, (announced) course changes may be reversed and those responsible may even be voted out of office. In foreign policy, too, public opinion provides such boundaries for decision-makers. Granted, leaders can usually move quite freely within these boundaries and lead public opinion on specific policy issues (Rattinger 1985;Saunders 2014). However, if the boundaries are overstepped and public interest activated, foreign policy issues can influence citizens' voting decisions (Aldrich et al. 2006;Schoen 2011). To avoid jeopardizing their popularity ratings, governments pay attention to public opinion between elections and may take it into account in their decisions (Baum and Potter 2015;Mello 2022;Sobel 2001). A comprehensive assessment of the Zeitenwende claim or, more generally, of the impact of Russia's invasion of Ukraine on German foreign policy therefore requires an analysis of public opinion.
Surveys of the German population conducted shortly after the start of the Russian invasion showed that sympathy ratings for Putin had fallen and the perceived threat from Russia had risen sharply (Graf 2022). There was also an increased willingness to defend allies in the event of a Russian attack, although this was still not a majority position (Smith 2022). Increasing military spending, sanctions against Russia, and arms deliveries to Ukraine met with majority approval (Bunde and Lubbock 2022;de Vries and Hofmann 2022). These early post-invasion surveys hence stand to some extent in contrast to traditional findings of German public opinion as "Venutian" rather than "Marsian" (Kagan 2004), as preferring cooperation over confrontation and persuasion over coercion (Duffield 1998;Gravelle et al. 2017;Rattinger et al. 2016). They suggest a shift from restraint to a more assertive posture. Against this backdrop, a constraint that may have prevented German policymakers from assuming such a posture in the past might be loosening.
However, caution is warranted in rashly diagnosing a Zeitenwende at the level of German public opinion. Previous major events, such as the terrorist attacks of September 11, 2001, and the Iraq war in 2003, influenced public opinion in Germany only with respect to a limited number of issues and for a limited time (e.g., Rattinger et al. 2016). It is therefore advisable not simply to assert a "new mindset in German society"  but to subject the question to careful empirical analysis.
Such an analysis should start with a specification of the key concept. The Zeitenwende claim can be usefully clarified by drawing on existing scholarly work on turning points (Abbott 2001) and related concepts (Capoccia and Kelemen 2007;Baumgartner et al. 2009). Based on this literature, we argue that a Zeitenwende claim at the level of public opinion requires significant and durable attitude change. This conceptual clarification is useful in guiding systematic analysis and evaluating existing evidence. For example, the shifts in public opinion described above, which were measured shortly after the start of the Russian invasion, cannot speak to the criterion of durability, and are hence not sufficient evidence for diagnosing a turning point.
In this article we take several steps toward answering the question of whether Russia's invasion of Ukraine marked a Zeitenwende at the level of German public opinion. First, we elaborate on significance and durability as criteria of turning points and specify these for public opinion as the unit of analysis. The main argument is that by distinguishing between different types of orientations-policy attitudes vs. core postures; attitudes thematically related to Russia's invasion vs. more removed-we can assess the likelihood and significance of the change that has occurred and, on that basis, make educated guesses about the durability of the change, even though not much time may have passed since the event. Second, we analyze German public opinion data from a panel survey conducted in the framework of the German Longitudinal Election Study (e.g., GLES 2021). These data were collected before (December 2021 and earlier) and after (May and October 2022) Russia's invasion of Ukraine, and the broad set of foreign and defense orientations that were repeatedly measured in this survey allow a detailed account of German mass belief systems in this domain. Furthermore, the panel data structure allows analysis not only of aggregate but also of individual-level change, hence providing a much more detailed account of whose opinion changed in what direction.
To anticipate the most important result, which will be elaborated upon in the following pages: the data suggest that Russia's invasion of Ukraine was not a historic turning point for German public opinion. Significant change was limited to attitudes toward policy issues and actors directly related to the Ukraine war. Neither did core postures change appreciably, nor did attitudes toward thematically more distant objects. In short, systematic analysis of survey data consisting of several pre-and post-invasion waves spanning a period of several years and extending into October 2022 shows that Germans did not see the world much differently after Russia's invasion of Ukraine than before.

Conceptual and Methodological Issues
We draw on the concept of turning point (Abbott 2001) to link the popular Zeitenwende claim to previous scholarly work and to guide our analysis of opinion change. Focusing on the analysis of life courses, Abbott defines turning points as short, radical shifts that set the life course on a new trajectory (Abbott 2001, p. 243). 1 We focus here on two aspects of this definition, each necessary but not sufficient to speak of a turning point. First, turning points are characterized by changes of a certain significance. It is not enough for Russia's invasion to change the world-or, in our case, public opinion in Germany-in some small, insignificant way; to be considered a turning point, the effects must be of fundamental importance. Second, turning points separate periods of stability. "What makes a turning point a turning 1 Similar concepts have been developed elsewhere, for example critical juncture (Capoccia and Kelemen 2007) and punctuated equilibrium (Baumgartner et al. 2009), in the literature on institutional change. These may prove particularly useful for scrutinizing the Zeitenwende claim at the institutional level. point rather than a minor ripple is the passage of sufficient time on 'the new course' such that it becomes clear that the direction has indeed been changed" (Abbott 2001, p. 245). Changes must therefore have a certain durability. If public opinion were to quickly return to its pre-invasion level, the event would not qualify as a turning point (for public opinion).
These two criteria are simple but also abstract-they need to be specified in order to make them accessible to empirical analysis. To specify the criterion of significant change in public opinion, we argue that not only the magnitude of change matters-more on that below-but also which type of attitude is affected. To develop this argument, we draw on the concept of belief system (Converse 1964). Accordingly, attitudes are organized in associative networks, where a horizontal and a vertical dimension can be distinguished (Converse 1964;Peffley and Hurwitz 1985). The horizontal dimension captures the idea that attitudes refer to different topics, and that associations between two attitudes on the same topic tend to be stronger than associations between two attitudes on different topics. Because of this pattern of association strengths, attitude change is more likely to spread among attitudes on the same topic and less likely to spread to attitudes on a different topic. The vertical dimension captures the idea that some elements of the belief system are more central, more important to the individual than others, which imposes a hierarchical structure on the belief system (Peffley and Hurwitz 1985). With respect to belief system change, elements at higher levels of the hierarchy are assumed to remain stable across situations, whereas elements at lower levels are more variable, as they are shaped by an interaction of the former and situational features.
Applying these concepts to the foreign policy realm, Hurwitz and Peffley (1987) argued that policy attitudes are typically situated at the lowest level of the hierarchy, whereas core postures are located at a higher level. While the former refer to the evaluation of specific policy options, the latter refer to general principles to which policy in a particular area should conform. Empirical research has fleshed out this idea, showing that citizens in many countries rely on three different postures when thinking about the foreign and defense policy domain (e.g., Chittick et al. 1995;Gravelle et al. 2017;Holsti and Rosenau 1990;Mader 2015;Wittkopf 1990). First, there are postures toward the extent of international involvement, with internationalists favoring an active role of their country in international affairs and isolationists favoring abstention. Second, individuals have principled positions on using military force-some reject it as a tool in international politics altogether, whereas others see its use as legitimate and effective. Finally, postures toward cooperation with other countries can be arrayed on a continuum ranging from multilateralists who prefer cooperation to unilateralists who prefer their country to act alone. In Europe and, in particular, Germany, postures toward the relationship with the United States also play an important role. While Atlanticists prefer their government to steer a close course to the United States and accept its leadership role in international affairs, others reject these notions to the extent of exhibiting a staunch anti-Americanism (Asmus et al. 2005;Katzenstein and Keohane 2007;Mader 2016).
Going back to the significance criterion, we use these prior insights on the structure of belief systems to argue that change in postures is more significant than change in policy attitudes and change in attitudes toward objects unrelated to Rus-sia's invasion of Ukraine is more significant than change in attitudes directly related to the event. The key idea, in other words, is to consider how far changes that occur at one point ripple through the belief system. The further away from the "point of impact" changes occur in both horizontal and vertical (upward) directions, the more significant the change is overall.
So far, we have stressed that significant attitude change is not exclusively related to the magnitude of the change-it is also relevant where in the attitude system changes occur. Nevertheless, the magnitude of change should not be ignored. A tiny change, even at the level of postures, is not significant in the sense of the above definition of turning points. But how to decide whether a given observed change is sufficiently large to be considered significant? Ideally, such a threshold could be derived from theoretical considerations. However, it is not clear to us what these might be. We are not alone in this, as existing research has no useful suggestions to offer either. Our threshold must certainly be above Page and Shapiro's (1992) of six percentage points, since their only concern was to distinguish genuine attitude change from random fluctuations (rather than detect significant change). 2 We tentatively suggest a threshold of ten percentage points, which corresponds, under reasonable assumptions, with a Cohen's d of 0.20 and hence indicates, by convention, a "weak effect" (Cohen 1988). 3 While this low threshold may seem undemanding and bias the analysis in favor of finding turning points, we will show that most opinion changes triggered by Russia's invasion of Ukraine do not even clear this hurdle. Given our overall conclusion that there has been no Zeitenwende at the level of German public opinion, relying on this low threshold hence represents a conservative strategy.
Turning to the criterion of durability, the obvious question is how long the change must persist to be considered "durable." We could retreat to the position that the more durable the change, the more legitimate it is to speak of a turning point, but that would be quite unsatisfactory. While setting any threshold is arbitrary to some extent, we believe it is appropriate to set it in the order of years rather than weeks or months. If a shorter period were used, short-term fluctuations could be misinterpreted as a turning point. Furthermore, turning points do not require immediate change, but may involve periods of change as long as that period is relatively small compared to the longer periods of stability around it (Abbott 2001: 251-2). Thus, if it takes some time for the public to understand the implications of Russia's invasion of Ukraine, it will not be possible to diagnose whether there has been a turning point for several years.
In terms of post-invasion data, our data analysis below covers two points in time, three months and eight months after the invasion, respectively. Accordingly, our analysis provides a solid basis for testing whether some initial changes took place and whether they survived another five months or disappeared in the meantime. But as the preceding discussion has shown, our analysis will still not be sufficient to settle the durability issue. If we find that public opinion has changed significantly, the question is whether this change will endure. If we find that public opinion has not changed significantly, the question is whether this can still happen (and how long the change will endure afterwards). We will return to these issues in the conclusion.

Boundary Conditions, Hypotheses, and Research Questions
Next, we briefly describe relevant boundary conditions of the empirical case to formulate hypotheses and research questions about attitude change. Due to the multitude of aspects that could be mentioned here, such a description invites selection bias and ex-post rationalizations. In order to avoid this, we have tied our hands by preregistering all hypotheses and research questions with the Open Science Framework (OSF) prior to accessing the data. 4 In general, the actions of elites and public discourse following Russia's invasion of Ukraine-which significantly shape the information environment in which citizens form opinions on political issues-are quickly described. The overwhelming majority of German and other Western elites condemned Russia's actions, called for solidarity with Ukraine and Western unity, emphasized the need for jointly upholding the principle of territorial integrity and liberal international order, and called for increased national defense efforts; in terms of policy, direct participation in the war by the country's own forces was ruled out, but significant assistance was provided to Ukraine, including arms deliveries, and strict sanctions were imposed on Russia (e.g., Adams 2022; Balfour et al. 2022;European Council 2022;NATO 2022). Assuming that public opinion reflected these mainstream interpretations in the aggregate, we should observe that public opinion after the invasion was more assertive in its policy orientations (H1) and more Western oriented (H2). 5 Given this assumption, we should also expect the public to be more multilateralist (H3) and less isolationist (H4) on average than before. Following the idea that belief systems are constrained on a vertical and horizontal axis, we furthermore expect that changes in postures are smaller than changes in policy attitudes (H5) and changes in policy attitudes to change less the more thematically distant they are from the topic of Russia's invasion of Ukraine (H6).
While these fundamental positions were largely shared across party lines in Germany, there were differences regarding more specific issues. For example, there were different levels of condemnation of Russia and different positions on how assertive the new Russia policy should be (Horowitz 2022;Pfaff 2022;Becker et al. 2022;Nienaber 2022;Feldenkirchen et al. 2022). This variation in elite cues opens the room for heterogenous opinion change along partisan lines (Campbell et al. 1960;Zaller 1992). Other individual differences may also lead to heterogenous reactions. Political involvement is a likely candidate. Given its effects on the willingness and capability of citizens to process information (e.g., Fazio 1990), those involved in politics might react differently to events than their fellow citizens who follow events less closely. We therefore conduct exploratory subgroup analyses to determine the extent to which the diagnosis for aggregate opinion change needs to be differentiated. Focusing on heterogeneity based on partisan loyalties and political involvement, we generally ask whether attitude changes differed across subgroups (RQ1) and whether differences led to a convergence or polarization of public opinion (RQ2). 6

Research Design, Data, and Measures
We analyze attitude change using a multi-wave panel survey. Two panel waves were conducted shortly before and after Russia's invasion of Ukraine (in December 2021 and May 2022, respectively). Differences in attitudes measured in these two surveys can be attributed to this event, since it was the only event related to foreign and defense policy that occurred in the short inter-wave period. This analysis hence allows evaluation of the significance criterion of turning points.
There are two extensions to this core element of the research design. First, we also consider pre-invasion data collected at various points in the 2017-2022 period. That way we can gauge change in a larger set of attitudes, even if causal attribution of differences between pre-and post-invasion data becomes more difficult. Principally, as the time between datapoints increases, so does the number of potential causes of attitude change. While we cannot discard this possibility, we are not aware of events or developments in that period that might have caused attitude change in this domain, and hence we deem it worthwhile to look at these additional data. However, to allow the reader to consider this potential inference problem when interpreting 6 We do not claim that this takes into account all possible sources of heterogeneity. Different socialization experiences during and after the Cold War could fuel region-and age-specific responses (e.g., Steinbrecher 2022); more generally, strong prior attitudes might color how citizens react to international events (e.g., Herrmann 2017). Supplementary Material S5 provides subgroup results for East and West Germans, Supplementary Material S6 documents heterogeneity in rates of change based on age and prior foreign and security policy attitudes. These additional analyses reveal some differences, but they are of small magnitude. Perhaps most notably, the change in support for certain assertive foreign policy measures appears to be weaker among East Germans than among West Germans. Overall, however, the additional analyses support our conclusion of a relatively homogeneous (non-)reaction of German public opinion, which is documented in detail below.  the results, we will indicate which time interval each comparison is based on when presenting the results. Second, we consider a second post-invasion datapoint (from October 2022) to provide a preliminary analysis of the fluidity of attitudes in the postinvasion period. This allows us to gain an impression of whether attitude changes in response to the invasion subsequently remained stable at the new level, whether changes continued in a given direction, or whether a trend reversal occurred. The data source is waves of a panel survey conducted in the framework of the German Longitudinal Election Study (e.g., GLES 2021). Questions on foreign and defense policy were included throughout the 2017-2022 period. Table 1 provides information about coverage of key items in different waves and the timing and sample sizes of these waves. The sample was recruited from the Respondi AG online access panel of pre-recruited persons using quotas that are representative of the German online population. Due to panel mortality on the one hand and sample refreshment between panel waves on the other, the size and composition of the sample varies across the analyses. 7 To maximize statistical power, we conduct all analyses with the largest possible sample, even though this might marginally affect the comparability of the findings. The analysis of intra-individual change between waves 21 and 22, for example, is based on about 10,500 observations; in the subgroup analyses, the results for supporters of The Left-currently the smallest party represented in the Bundestag-are based on about 1000 respondents. In addition to the large number of foreign and defense policy questions repeatedly asked of the same individuals, these large sample sizes are a unique feature of this data source.
Corresponding to our theoretical discussion, we analyze opinions on different topics and at different hierarchical levels of the belief system. Needless to say, the selection of available items is not perfectly suited to the purpose of this specific analysis, as the Russian invasion of Ukraine and the salient policy issues arising from it were unforeseeable, and tailored items in the pre-invasion waves hence unavailable. For example, attitudes toward arms deliveries, sanctions, and compliance with alliance commitments were not measured in the waves preceding the invasion and thus cannot be included here. Nevertheless, the pool of available items does allow analysis of all types of orientations discussed above. We provide an overview of the survey items used here and document the original question wording and response options in Supplementary Material S1.
Several items refer to policy issues more or less directly related to the Ukraine war. These are (1) increasing German military spending, (2) the preliminary acceptance of Russia's annexation of Crimea, (3) whether German governments should seek good relations with Putin, and (4) whether Germany should rely less on cooperation and more on confrontation in dealing with Russia. With regard to issues not directly related to the invasion, questions about a confrontational (rather than cooperative) China policy and stronger EU efforts to protect its external borders are available. The former is particularly important for our endeavors, as it is completely analogous to the question about a confrontational Russia policy but differs in topical closeness to Russia's invasion of Ukraine. These items are hence particularly suited to address hypothesis 6 (i.e., the expectation that the further away attitudes are thematically from Russia's invasion of Ukraine, the less they changed). Attitudes toward these issues were all measured using a five-point Likert-type scale (strongly disagree-strongly agree).
We also examine views of foreign actors directly (Russia) or indirectly (U.S. and China) involved in the war, even if their hierarchical position in the belief system is somewhat unclear. In part, they presumably reflect country stereotypes, which should be located at a higher level; at the same time, they strongly depend on contextual features-e.g., who occupies the White House strongly influences public opinion of the U.S. (Wike et al. 2021). We therefore assume that they are rather located on the lower level of the hierarchy. Views of foreign actors are measured with items from a battery that asks respondents what they think of different countries and politicians. Respondents could register their views on an 11-point scale (I don't think anything of them at all-I think very highly of them).
Foreign policy postures are captured with a set of established indicators (e.g., Graf 2020; Mader 2015; Steinbrecher 2018). For each posture, the responses to two items were aggregated into an additive index. The items are statements that express the respective foreign policy principle at an abstract level, in opposing wording to avoid approval bias. For the posture toward the use of military force, for example, the statements are "War is sometimes necessary to protect a country's interests" and "The use of military force is never justified" (reverse coded). Responses were recorded using a five-point Likert-type scale (strongly disagree-strongly agree). For the items used for the remaining postures, we refer again to Supplementary Material S1.
The correspondence between these orientations and the hypotheses is as follows: all policy items address the assertiveness dimension, as does the core posture on the use of military force (H1). A Western orientation is expressed in positive views of the U.S. and an Atlanticist posture (H2). Hypotheses H3 and H4 can be directly tied to the corresponding postures, multilateralism, and isolationism.

Results 8
Is there evidence that Russia's invasion of Ukraine triggered aggregate opinion change in the German population? To answer this question, we look at the percentage of agreement with a given survey item. Responses to the military spending item, for example, were transformed into a dummy variable indicating whether respondents chose one of the two agreement options ("agree" or "completely agree") or one of the three other substantive options ("completely disagree," "disagree," "neither agree nor disagree"). Rare refusals to answer were treated as missing values. The 11-point scales capturing respondents' views of Russia, the U.S., and China were treated analogously; thus, we report the percentage of respondents who expressed a positive opinion about each actor (by choosing a response option above the middle category). Finally, the sum indices capturing postures were also dichotomized above the midpoint of the scale, so that in each case the percentage is reported at which respondents (tend to) agree with the use of military means and a multilateralist, isolationist, and Atlanticist foreign policy, respectively. 9 Fig. 1 reports these data from multiple pre-and post-invasion waves of the panel survey. Data from the last wave before the invasion are colored solid red; data from the first wave thereafter are colored solid blue; data from waves further away in time are displayed in progressively lighter red or blue. Figure 1 shows that German public opinion became more assertive on policy issues directly related the Russian invasion of Ukraine. Comparing the two datapoints bracketing the invasion, support for higher military spending increased significantly (+26 percentage points), as did opposition to the annexation of Crimea (+25) and support for a more confrontational Russia policy (+16). Striving for good relations with Putin became significantly less popular (-23 points). In contrast, there was no or only minor change in all other orientations considered here-core postures, attitudes toward policy issues not directly related to the war, and views of Russia, the U.S., and China. 10 Considering the second datapoint after the invasion, from October 2022 (in light blue), shows subsequent trends. First, there is no evidence of a delayed reaction to the invasion. As the war in Ukraine continued through the summer and fall of 2022, core postures and attitudes toward thematically distant policy issues and actors 8 Reproduction materials for all results reported in this article are available at Harvard Dataverse: https:// doi.org/10.7910/DVN/WY2OFU. 9 Another option is to treat the scales as metric and look at changes in means. We provide these in Supplementary Material S2. 10 The (relative) stability of the postures is not a methodological artifact of using two-item indexes. Supplementary Material S3 reports the stability of responses to the individual items, which is also high. The minor change in views of Russia, however, is partially an artifact of the dichotomized measure. The intraindividual analysis below shows that there was actually significant movement toward more negative views of Russia.

Views of actors
Policy attitudes Fig. 1 Stability and change in aggregate opinion. Reported are percent of respondents who score above the midpoint of the original scale. December 2021 data in solid red (last wave before Russia's invasion of Ukraine), May 2022 data in solid blue (first wave after the invasion); data from more distant waves in increasingly transparent red and blue, respectively. See Supplementary Material S2 for data in tabular form. Pre-invasion marker for "Confrontational CHI policy" completely overlaps with post-invasion marker remained at pre-invasion levels. Second, the changes in attitudes diagnosed above prove stable in some cases, but in others the initially more assertive stance seems to be moving back to the conciliatory pre-invasion levels. While greater support for higher defense spending and rejection of recognition of Russia's annexation of Crimea prove durable, there were again more respondents who wanted good relations with Putin and a Russia policy that focuses more on cooperation than confrontation.
With respect to hypotheses H1-H4, we can summarize that the direction of the change observed immediately after the invasion was as expected. If we consider the magnitude of change, however, H2-H4 have to be rejected completely and H1 can be only partially confirmed. There was a shift toward greater assertiveness on certain policy issues, but this did not extend to issues further removed from the Ukraine war or to foreign policy postures. Support for a more confrontational China policy was virtually identical before and after the Russian invasion, and the increase in the percentage of citizens generally supporting the use of military force (from 18 to 20%) is so small as to be negligible. Similarly, favorable views of the U.S. (+1 percentage point) and Atlanticism (+5) were more common after the invasion, but not to the extent that one could speak of a significant spread of "Western orientations" (H2); multilateralism became more widespread among the German population (+4) and isolationist attitudes declined (-5), but again the magnitude of these changes is too small to constitute a significant change (H3, H4).
Hypotheses H5 and H6 perform much better. We find movement mainly in attitudes toward political issues directly related to the invasion of Russia, while there is little change in postures and attitudes toward issues that are thematically distant. As noted above, these hypotheses are key to gauging the significance of belief system change overall, and hence to understanding whether the Russian invasion marked a turning point for German public opinion. The aggregate-level findings presented so far give a clear answer: changes are limited to a small topical area at a low hierarchical level of citizens' belief systems; they do not extend beyond that either in the horizontal dimension (to thematically more distant policy issues) nor the vertical dimension (to core postures). Based on the criteria we have developed above, there is thus insufficient evidence overall for a turning point at the level of German public opinion.
Before going further, we should revisit the concern that causal attribution is not entirely unproblematic in a design of this type. Perhaps the observed patterns of stability and change are random or have causes other than the Russian invasion of Ukraine. This concern is especially important in cases where the pre-invasion datapoint is further in the past. Arguing against much empirical relevance of this objection is the fact that all orientations considered here-even those that changed in the course of the invasion-were stable over a long period before the invasion (see Supplementary Material S2). Thus, a long uneventful period was associated with stability in public opinion, while a salient event was associated with (some) change. This pattern encourages us to interpret the differences that appear in the before/after comparison as actually being reactions to the event.
The above diagnosis of extensive opinion stability is based on aggregate findings. It is possible that the stability in the aggregate masks significant, countervailing change at the individual level that cancel each other out when aggregated. To test this possibility, we first make use of the panel data and examine the extent to which the opinions of individual citizens have changed in reaction to the invasion. We calculated between-wave differences (first post-invasion minus last pre-invasion datapoint) using the original scales and trichotomized these differences. Table 2 contrasts respondents who exhibit change of +/-1 scale points or less with respondents who exhibit a more than +/-1 scale-point change. We believe these to be reasonable cut-off values to discriminate between stability and change. We do not use an even more rigorous method because the change in a single scale point could reflect random variation. 11 Furthermore, it should be noted that the comparability of results is attenuated by the use of different response scales. The use of an 11-point response scale is more likely to result in response variability than use of a five-point scale for methodological reasons alone, whereas the construction of indexes to capture postures increases the stability in the measure. 12 Table 2 shows a similar pattern at the individual level as was shown before at the aggregate level. Policy attitudes toward issues not directly related to the Ukraine war were essentially stable-regarding the question of how to handle China, for example, four out of five respondents gave the same answer or answered differently by only one scale point. Those who did indicate a change in opinion did so in roughly equal proportions in both directions. A similar result emerges for views of the U.S. and the core postures, although here the absolute stability levels are somewhat different, partly for the methodological reasons mentioned above.
At this point, it is instructive to take a closer look at the views of foreign actors. More clearly than above, we see a significant net shift toward more negative views Fig. 2 Post-invasion trajectories of change. Reported are group sizes with different types of reaction to the invasion (increased support vs. no change) and the subsequent post-invasion change within these groups, with the total sample as the percentage base. Results for the third type of reaction in the first interval (decreased support) not shown due to small group sizes. See Supplementary Material S4 for complete results of Russia. More than a third of respondents expressed a more negative opinion, while only seven percent expressed a more positive opinion. By comparison, views of the U.S. and China are far more stable, and among those who changed their opinion, there is no clear trend in one direction or the other. Thus, a similar pattern emerges here as with political attitudes: changes are limited to views about Russia, the country directly involved in the war.
The post-invasion trajectories of individual-level change parallel those of the aggregate analysis and are therefore documented here only by way of example. Figure 2 shows for attitudes toward increasing military spending that the shift also endured at the individual level. In October 2022, 24% of respondents still indicated increased support for higher military spending (relative to pre-invasion levels)-almost as many as in May 2022, when the value was 27%. In contrast, with regard to the question of what Germany's policy toward Russia should look like, there was a certain trend back toward support for a cooperative policy, as the aggregate analysis already showed. While 20% in May 2022 exhibited change toward a more confrontational Russia policy compared with before the invasion, only 12% did so in October 2022.
The second step in our analysis is to look at selected subgroups. Even if the changes in the aggregate are not sufficiently large to diagnose a turning point, this could well be the case in subgroups. A follow-up question concerns the convergence of public opinion-it is possible that different rates of change led Germans to view the world more similarly after the Russian invasion than before. The subgroup analysis was conducted for all orientations for which data are available and our conclusions are based on this overall picture, but we present only a selection of results here. This is done for lack of space, for the sake of clarity, and to avoid repetition-the subgroup differences are quite similar across the different (types of) orientations, so there is no value added by showing all results here. All results are provided in Supplementary Material S5, alongside regressions that model difference variables as a function of subgroup variables.
We find that the differences between the subgroups are by and large small (RQ1). In other words, we find no evidence that there has been a turning point in the opinions of at least certain subgroups. This is true in particular for the foreign policy Pre-and post-war opinions, by party group and political interest. Reported are percent of respondents within subgroups who score above the midpoint of the original scale. Pre-invasion data refers to the last available data before the invasion; post-invasion data from 05/2022. The red vertical line is the preinvasion average, the blue line is the post-invasion average postures-their stability is high without exception in all subgroups. Where we do find differences, they are gradual rather than in kind, i.e., the changes are going in the same direction but at slightly different rates. Specifically, attitudes tended to change more among supporters of mainstream parties than among supporters of fringe parties and citizens with no party affiliation, and changes were somewhat more pronounced among citizens with high political interest than among citizens with low political interest.
The most interesting subgroup results are displayed in Fig. 3 and relate to support for increased military spending and a confrontational Russia policy (top row), and to views of Russia and the U.S. (bottom row). They illustrate the general patterns described in the previous paragraph well, but they are unrepresentative in terms of magnitude of change overall and subgroup differences-change and subgroup differences are smaller in the other cases.
Consider defense spending (top-left plot). There were significant increases in support for more defense spending in all subgroups, but the rate of change varied between supporters of the mainstream parties (27-36 percentage points) and those of AfD and The Left (13 and 17 points, respectively). Analogously, increases in support for a more confrontational Russia policy were also more pronounced among the former than the latter. These patterns are in line with party positioning. While the parties on the political fringes condemned the Russian attack, they did so less vehemently and were critical of German arms deliveries to Ukraine. They were also quick to provide justifications for Russia's attack and call for diplomatic solutions, which at the time would have meant accepting Russian territorial gains, if not Ukrainian loss of sovereignty (Horowitz 2022;Pfaff 2022).
The differences in change among supporters of the established parties are small and essentially negligible. Here, partisan disputes over the Russian invasion apparently did not spread to their followers. For example, one might have suspected that the SPD's supporters moved less, given Chancellor Scholz's reluctance in delivering arms to Ukraine (Nienaber 2022). The Greens called for a more decisive approach and, in the process, surprisingly quietly moved further away from their pacifist roots ). Accordingly, one might have expected supporters of the Greens to move the most on policy issues, and also to show a comparatively large change in their postures toward the use of military force. However, we see no such differences. The differences in positioning in the day-to-day business of party politics were apparently not sufficiently fundamental to leave a lasting impression at the public level.
Finally, we can answer the question of whether Russia's invasion of Ukraine led to a convergence or polarization of public opinion (RQ2). Since subgroup differences in attitude change were generally small, the straightforward answer is that there was neither. Only at second glance can trends of polarization be discovered for some issues, while for others there is a slight convergence. Views of the U.S. is the only case with an instance of subgroups changing in opposite directions. Supporters of the two fringe parties and citizens with no party affiliation-who were more critical of the U.S. from the start-became more critical, whereas the views of supporters of the established parties became more positive. As Fig. 3 shows, however, these opposing effects were substantively insignificant. In addition, there were some cases in which, depending on the distribution of opinion before the war, different rates of change led to polarization (e.g., in military spending) or convergence (e.g., in views about Russia). Overall, however, these effects are too small to call into question the general finding that neither polarization nor convergence occurred.

Conclusion
Our analysis suggests that Russia's invasion of Ukraine passed by the German public without leaving deep traces. This diagnosis is based on a simple conceptual idea and clear empirical findings. Conceptually, we propose that attitude change should be considered significant only if the invasion caused widespread changes of a certain magnitude in mass belief systems. We posit that change in postures is more significant than change in policy attitudes, and change in attitudes not directly related to an event is more significant than change in attitudes that are directly related to it. Empirically, we find that policy attitudes toward war-related issues such as military spending and a confrontational Russia policy became more assertive, but that neither postures nor attitudes toward issues (and actors) that are not directly related to the invasion changed. This pattern of limited, topical change and more general, overarching stability holds regardless of political orientation and interest. Moreover, in the post-invasion period studied, there is nothing to suggest a continuing shift in orientations that might eventually add up to significant change. In short, if we apply our criteria for significant attitudinal change, we conclude that there has been no such change, and thus no Zeitenwende at the level of German public opinion (yet).
The diagnosis of course depends on the criteria used. We have already pointed out above that the chosen threshold of ten percentage points for a change that is "significant" in terms of magnitude is somewhat arbitrary. However, we also argued that this threshold is set rather low-setting it higher leads even more clearly to the diagnosis made here. Furthermore, conceptual decisions could reasonably be made differently. For example, we have categorized attitudes toward defense spending as policy attitudes, following convention (Hurwitz and Peffley 1987;Wlezien 1995). If these were interpreted as indicating general postures toward the role of the military in foreign policy instead, the results on the lack of opinion change at higher levels of German belief systems would be more mixed. Concrete political issues can indeed become powerful symbols, and public attitudes toward these issues can in turn occupy central positions in mass belief systems (Sears 1993). Given the importance of this issue in the German Zeitenwende debate and in light of the state of the German armed forces, it is not impossible that military spending will acquire such symbolism. This reading conflicts, however, with existing evidence of the context dependence of these attitudes (Wlezien 1995), and even if we assigned them posturelike status, the large change in this one case would be offset by the evidence of stability of the other postures. Even then, a balanced assessment of all the available evidence seems to amount to rejecting the thesis of a turning point at the level of public opinion.
Beyond the issue of diagnosing a turning point in public opinion, the results presented here are in line with more general insights about citizens' foreign and defense policy attitudes. First, we find no evidence that these opinions are particularly fickle or in any way not "real." While this was the verdict of early observers of public opinion on foreign and defense policy, subsequent studies have revised this view (see Holsti 1992). Our findings strike the same chord. The changes in opinion that we do find are quite compatible with notions of a rational public (Page and Shapiro 1992) that reacts like a thermostat to changes in its information environment (Wlezien 1995). We also find exactly the pattern in core postures that one would theoretically expect-they turn out to be stable across contexts. Taking the theory of posture-based attitude formation seriously (Feldman 1988;Hurwitz and Peffley 1987), we should expect changes in policy attitudes to be short lived without changes in overarching postures. When the salience of the Ukraine war fades and other issues capture the citizens' attention, postures will remain as central determinants of policy attitudes. Citizens who, under the impressions of the war and the (largely) consensual reactions of the established political parties, were ready to support assertive policies "against type" may then return to their default, posturebased position. What assertiveness German public opinion did show in the months after the invasion could then turn back into restraint.
This brings us to the policy implications. At present, there seems to be no mandate for a fundamental change in Germany's foreign and security policy orientation. Depending on the normative point of view and whether one views the Ukraine war as a turning point for international relations, German public opinion thus appears as an anchor of stability in stormy times or as an obstacle to urgent changes in the status quo of Germany's strategic posture. To the extent that a fundamental reorientation has occurred at the level of German policymakers (Bunde 2022), the stability at the public level creates-or rather deepens (Oppermann 2019)-a disconnect that may lead to tensions in the future. On the other hand, the electoral incentives from stability in public opinion make fundamental policy change a risky, even unreasonable strategy from a political point of view. Therefore, any pressure for change emanating from the international situation could soon be neutralized by domestic, electoral considerations.
Our conclusions are of course subject to the proviso that significant changes in public opinion will not still occur. Here lies an obvious weakness of the present empirical analysis: our post-invasion data cover only the period until October 2022, only about eight months after the event. We are thus unable to examine the durability of change, the second criterion for a turning point that we discussed above. Perhaps significant change in public opinion will yet occur. Triggers could be a single prominent event or a series of smaller events that cumulatively lead to such change. Beyond academic questions about what constitutes an event and whether attitude change would then still be a reaction to the Russian invasion, the main issue here is what exactly this other event (sequence) might be and how likely it is to occur. In terms of singular evens, candidates might be the Russian use of weapons of mass destruction, complete Ukrainian defeat, an extension of the war to other (NATO) countries, or severe economic shocks on the home front. We do not want to speculate how likely these or other potentially relevant events are, but merely note that we would assume an inverse relationship of the probabilities in which they occur and in which they cause a significant shift in attitudes.
As for smaller events that could cumulatively lead to significant change, sustained opinion leadership could convince the public of new foreign and security policy imperatives (Giegerich and Terhalle 2021). We suspect, however, that it will be difficult even for committed policymakers to achieve this-at least in the short term and insofar as it involves changing citizens' core postures. Such opinion leadership would likely require a sustained elite consensus (among the established parties) that grapples with difficult value trade-offs and defends these positions against other pressing concerns. Whether such a consensus will emerge among German elites and whether serious attempts will be made to bring the public along remains uncertain for the time being. It is also uncertain whether such an attempt at opinion leadership would be successful. Research to date says little about the potential of elite communication for shaping core postures. Systematic analyses of the malleability of these orientations-beyond the results presented here-are not yet available and are an important task for future research.
Prudence dictates that no conclusive assessment of the German public's reaction of Russia's invasion of Ukraine be made on the basis of the evidence presented here. This much, however, is clear: the invasion did not immediately represent a Zeitenwende for German public opinion on foreign and defense policy. Whether this early assessment will have to be revised and whether turning points occurred in other domains, the future and further research will show.
Funding Open Access funding enabled and organized by Projekt DEAL.

Conflict of interest M. Mader and H. Schoen declare that they have no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4. 0/.