Utility-based predictions of military escalation: Why experts forecasted Russia would not invade Ukraine

ABSTRACT

When Russia amassed troops in the winter of 2021–2022, many analysts deemed a large-scale invasion of Ukraine unlikely. Surveying the expert literature, we establish that these arguments largely relied on utility-based reasoning: Analysts thought an invasion was improbable, as it would foreseeably entail massive costs for Russia, its people, and its regime. We show that this regnant expert opinion had not sufficiently accounted for the Russian regime’s tendencies to increasingly accept risks, coupled with an inadequate processing of information on Ukrainian and Western views and policies. We argue that analysts miscalculated partially because the most prominent facts, long-term trends, and causal mechanisms available to them jointly suggested Russian cost-sensitivity, but provided only weak signs of countervailing factors. We thereby showcase that good forecasting requires explicit theory with a view on multiple interacting causal factors, area expertise and Socratic humility on the extent, context and certainty of our findings.

KEYWORDS:

• Geopolitical forecasting
• rationalist explanations of war
• heuristics in international politics
• Russo-Ukrainian war
• deterrence
• interdependence

When Russian forces amassed near Ukrainian-held territory in late-2021, many, if not most expert predictions argued Russia would not all-out invade Ukraine.¹ Yet, it did. In this forum article, we ask: Why were the forecasts wrong?²

Our answer has four elements. First, we find that most of these forecasts relied on utility-based thinking, that is, arguments that assume actors choose between available policies by weighing their expected costs and benefits. Analysts argued that the Russian regime had plenty of information available that suggested an invasion would likely lead to protracted fighting, costs, international isolation, and sanctions. As these likely consequences would ultimately undermine the sources of its domestic power, the forecasts expected the Russian regime to abstain from full-scale invasion.

Second, using post-invasion data we evaluate the performance of these utility-based predictions. We find that pre-invasion accounts rightfully predicted Russia would suffer massive costs if it attacked, and that it had access to plenty of information suggesting this. However, the forecasts underestimated how willing the Russian regime was to accept risks and how prone it was to miscalculate the risks and costs of an all-out invasion. We also hypothesize that other factors might have contributed to the invasion as well as to the failure to predict it, although they remain to be corroborated or disconfirmed.

Third, we argue that this failure of prediction³ was not due to inherent flaws in utility-based thinking. Rather, we find that analysts miscalculated partially because the most prominent facts, long-term trends, and causal mechanisms available to them jointly suggested Russian cost-sensitivity, but provided only weak signs of countervailing forces. We find that more successful forecasts considered multicausal developments and worked actively against the potential influence of so-called availability heuristics by reflecting on the known and unknown assumptions of their analyses.

Fourth, our results corroborate findings on the necessity and difficulty of combining theory and case-specific knowledge for political forecasts. We find that recommendations and predictions should reflect on their assumptions, and that these assumptions should be grounded in consistent and corroborated theorizing that is mindful of multicausality, interactions, and scope conditions. Lastly, we call for Socratic virtue: practicing and demanding humility about the extent of our knowledge, and insisting on it where we have it.

Some notes on our approach: We considered expert forecasts from late-2021 to the start of the invasion. As our interest lies in the most prominent (and arguably most consequential) expert opinions, we focused on English-language accounts, due to the prominent status of English in the relevant disciplines and among the major states now supporting Ukraine. We also considered some forecasts by Russian experts. “Expertise” is, of course, a subjective concept. We considered writers that we could reasonably group under one or more of the three following categories: theory-focused scholars working on conflict initiation and deterrence (mostly in academia), policy experts with a focus on war and military affairs (mostly at think tanks), and area experts with a focus on Eastern Europe and Russia in particular (academia and think tanks). As the invasion manifested itself over a few months, and as academic publications have a longer production cycle, short reports and media statements vastly outnumber peer-reviewed academic publications.

Lastly, we want to stress that we do not claim or imply that any of the analysts cited acted unscholarly or displayed bad craftmanship in their work. In fact, we find that many accounts were mindful about potential blind spots and cautioned against putting too much confidence into their predictions. We, too, had found a full-scale invasion unlikely, and we had also largely based this expectation on utility-based thinking.

Expert views before the invasion

Pre-invasion accounts of the likelihood of a Russian invasion were overwhelmingly arguing that an all-out invasion was improbable, and a great number of them, possibly a vast majority, explicitly argued that this was due to the enormous costs the Russian regime should expect to incur.

The structure of this argument is a basic application of rationalist models on deterrence and war (e.g., Mazarr et al., 2018): Presuming leaders seek to best reach their goals using available resources and information,⁴ they will weigh the expected benefits and costs of policies and choose whichever promises the greatest net-utility. Thus, we should generally expect leaders to rarely choose the extremely risky and costly option of offensive war (Fearon, 1995), especially when there are clear signs that victory would be difficult to achieve, occupation would be hard to manage, and other states could and would inflict prohibitive damage in response.

An abundance of forecasts advanced versions of this argument. For example, in a report for the Institute for the Study of War, Kagan et al. (2021) surveyed the “risks and costs Putin would have to expect” when ordering an invasion, which led them to “forecast that he is very unlikely to launch an invasion”. Similarly, in a joint op-ed for the Washington Post, Samuel Greene and Graeme Robertson (2021) argued that a radical shift in Russian policy toward all-out war was unlikely, as they attributed to Putin a tendency to stay clear of risks and costs that would be associated with such an undertaking. Both are pre-eminent Russia scholars, Greene being the director of the Russia Institute at King’s College, London and Robertson the director for the Center for Slavic, Eurasian and East European studies at University of North Carolina at Chapel Hill.

Others also named expectably high costs and risks as a reason to not expect an invasion. Among them were: Eugene Chausovsky (2021), senior analyst at the Newlines Institute; Mark Galeotti (2022), expert on Russian security policy and senior associate fellow of the Royal United Service Institute; Keir Giles (2021), Chatham House Senior Consulting Fellow; Sergei Guriev (2022), economics professor at Sciences Po, Paris; Jeff Hawn (2021), Russia and security researcher at the London School of Economics; Harald Malmgren (2022), former adviser to various U.S. presidents and political consultant; Steven Pifer (2021), former U.S. Ambassador to Ukraine and current Nonresident Senior Fellow at the Brookings Institution; Ivan Timofeev (2021), Director General of the Russian International Affairs Council; and Harun Yilmaz (2022), editor of the Routledge Central Asia Research Forum Series.

Many of these forecasts, as well as other analyses at the time, went into detail on the expectably massive costs that Russia would face if it invaded. The costs of fighting itself were a prominent theme in these accounts. For example, Arnold (2022) as well as Royal United Services Institute analysts Jack Watling and Nick Reynolds (2022) highlighted the various downsides Russia would likely face by trying to occupy, control, and govern large swaths of Ukrainian territory. Indeed, even the head of the Russian Academy of Science’s Center for Ukrainian Research Victor Mironenko (2021) agreed, deeming a war counterproductive and costly. He also stated that Ukraine would turn into “another Afghanistan” (referring to the Soviet-Afghan war from 1979 to 1989).

Most accounts that we surveyed explicitly argued that Russia would face a massive economic fallout, and many put it front and center. Among these were those by Kagan et al. (2021) and Chausovsky (2021) as well as Ed Arnold (2022) from the Royal United Service Institute; Henry E. Hale (2022), Russian public opinion specialist and professor at George Washington University; and Robert Hunter (2022), former U.S. Ambassador to NATO.

Russian vulnerabilities to tougher Western sanctions were particularly highlighted. Alexandra M. Vacroux (2022b), director of the Harvard Davis Center for Russian and Eurasian Studies stressed the severe short-term impact of banning Russia from the Society for Worldwide Interbank Financial Telecommunication (SWIFT). Emphasizing sanctions on Russian energy exports, a study for the Vienna Institute for International Economic Studies stated that the “Russian economy remains highly exposed in a number of dimensions, which would entail high costs with a severe sanctions regime between Russia and the EU and the United States” (Astrov et al., 2022; cf. Rodrigues Vieira, 2023, this issue). Atlantic Council senior advisor Harlan Ullman (2022) inferred that “[Putin] knows that the costs to Russia and to him personally will be high and possibly unaffordable. Sanctions and further isolation will hurt.”

The expectation of Russian cost-sensitivity is also reflected in experts’ recommendations to increase expectable costs to Russia by threatening sanctions. Among these were, for example, the director of the European Council on Foreign Relations Mark Leonard and Chairman of the Center for Liberal Strategies Ivan Krastev and Mark Leonard (2022) as well as a group of experts at the Center for Strategic and International Studies (Jones et al., 2022).

Various analysts drew a connection between these risks and costs on the one hand, and the political survival of the Putin regime on the other. For example, American Security Project Chief Operating Officer Herberto Limas-Villers and Matthew Wallin (2022) stated that economic sanctions, a protracted war, international isolation, and mounting casualties “could spell endless turmoil for a leader eager to remain in power.” Indeed, a major theme in the utility-based forecasts was that these expectable costs would undermine domestic support for the Russian regime and thus disincentivize aggression (cf. Greene & Robertson, 2021; Hale, 2022; Kagan et al., 2021; Vacroux, 2022a).

Missing the mark?

The reviewed utility-based estimates were right about two of their core contentions: First, through the invasion, Russia did indeed pay an enormous price in blood, treasure, goodwill, and reputation (cf. e.g., Garton Ash et al., 2023; Snegovaya et al., 2023; Stoner, 2022).

Second, the Russian regime had access to plenty of information that strongly suggested an invasion would entail these risks and costs. Even the comparatively minor Russian incursions into Ukraine since 2014 had resulted in costly Western sanctions (cf. Andermo & Kragh, 2021), economic and political support for Ukraine, and a much more determined and forward NATO policy toward Russia (cf. Driedger, 2021a). This new Western policy toward Russia remained largely in place up until early 2022, even during major political upheavals in key NATO member states, like the United Kingdom referendum to leave the European Union and the election of Donald J. Trump as President of the United States (Driedger, 2020, 2021a). Meanwhile, anti-Russian sentiment and government support had significantly increased in Ukraine, and Kyiv had invested heavily in its armed forces (Driedger, 2023, pp. 13–15). Prominent Russian security experts, for example Timofeev (2021, 2022) had stressed all these factors prior to the invasion.

What, then, did the forecasts miss? With the benefit of hindsight, we can corroborate two mutually reinforcing factors that jointly help to explain the Russian decision to decide on invasion despite the abundance of information on expectably massive costs.

The first is a long-term term increase in the Russian regime’s willingness to accept risks in the pursuit of its goals. When weighing options with the same estimated net-utility, risk-averse leaders prefer sure-thing outcomes over lottery outcomes. For risk-acceptant leaders, the reverse applies. Risk-neutral leaders prefer neither. When disputing a good with another party, risk-acceptant leaders are more likely to initiate war because they are more willing to accept two lotteries: one on the uncertain probability of the other side resisting their coercive policies (likelihood of war), and one on the uncertain extent of downsides of the war itself compared to a peaceful settlement (costs of war, see Tarar, 2022). Thus, if the Russian regime was indeed highly risk acceptant, the mere expectation of high risks would not deter it from an attack, provided it also expected that the likely benefits would be significant as well.

Indeed, Driedger (2023) found that, since the mid-2000s, the Russian regime’s risk acceptance in offensive military operations has grown continuously, reaching a peak with the 2022 invasion. In its planning and design of its attacks on Georgia (2008), Crimea (2014), Eastern Ukraine (2014–2021), and Ukraine as a whole (2022), the Russian regime increasingly accepted risks of international backlash, waning domestic support, and conflict entanglement to ensure the respective campaign’s goal (Driedger, 2023).

The second factor explaining the decision for invasion in the face of information suggesting massive costs is misplaced certainty. Risk acceptance causes escalation due to leaders being more willing to engage in the lottery represented by the uncertainties of war. In contrast, misplaced certainty causes escalation because leaders falsely assume that outcomes are more predictable, risks lesser, and rewards higher than they actually are. We would expect a particularly aggressive policy if the leaders are both highly risk acceptant and simultaneously underestimate the magnitude of expected risks and costs. Misplaced certainty usually stems from biases and bureaucratic dysfunctionalities at various levels of the state hierarchy (Mitzen & Schweller, 2011).

Since the very start of the invasion, analysts have pointed to misplaced certainty as a key explanation. According to this view, Russian elites had a more optimistic view on the likely risks and costs of an attack than were actually warranted, given the information available to them (e.g., Kormych, 2022; Kupchan, 2022). These misplaced certainties seem to stem from both leadership biases and bureaucratic deficiencies. Defected Russian state officials have since reported that, increasingly, a culture of ideologized kowtowing and affirmation of the presumed wishes and views of the power center replaced rational processing of intelligence (e.g., Bondarev, 2022; Karakulov, 2023). Director of the U.S. Central Intelligence Agency William J. Burns and director of U.S. National Intelligence Avril Haines have stated that, from their intelligence and personal conversations with Putin, the Russian president was skeptical of a forceful Western reaction, considering high energy prices in the winter and domestic changes in France and Germany (POLITICO, 2023). Western security officials have reported that even supposedly prominent Russian national security advisors were startled when they were told about views that Putin had relayed to Western counterparts (POLITICO, 2023).

One apparent Russian miscalculation regarded the relative strengths of the Russian and Ukrainian fighting forces. However, the accounts surveyed by us also did not find it likely that Ukraine could stop a Russian advance on Kyiv outright. Indeed, key security officials in both the United States and Russia had vastly overestimated Russian military power relative to Ukraine (cf. also Seitz, 2023, this issue). After the start of the fighting, wide gaps between Russian planning and the actual capabilities on the ground became evident (POLITICO, 2023). Furthermore, it quickly became clear that Russia had underestimated how united and determined Ukrainian society would rally around the cause of national defense (cf. Kudlenko, 2023).

Rather than on the result of a direct military clash, most utility-based forecasts had focused on the likely repercussions of Western sanctions. We now have evidence of misplaced certainties in this area. Indeed, key figures in the Russian regime seem to have overestimated European economic dependence on Russia and hence downgraded the risk of extensive sanctions and support for Ukraine by Western Europe. One sign of this is the Russian failure to move significant assets from Western institutions before the invasion, failing to protect them from punitive measures (Seddon & Ivanova, 2022). When Russian reserves were frozen right after the start of the invasion, Russian foreign minister Lavrov stated that “No one could predict before the war that the West would [actually] apply such [harsh] sanctions” (Lavrov, 2022). As this statement implies Russian intelligence failures and that the sanctions actually hurt the Russian economy, it arguably reflects genuine befuddlement by the Russian elite.

The regime might have inferred from previous Western actions that core sectors of the Russian economy would be spared after the invasion. Even though the sanctions since 2014 cost Russia losing an estimated 2.5–3 percent of annual gross domestic product (Åslund & Snegovaya, 2021), the crucial oil and gas sectors were largely exempted (Rutland, 2022). Indeed, only one year after the initial aggression in 2014, construction of the Nord Stream 2 pipeline project between Russia and Germany was started. When signs of a Russian military build-up mounted in late 2021/early 2022, German elites did not publicly threaten to cancel or freeze Nord Stream 2 in the event of an all-out invasion (Driedger, 2022). Furthermore, a leading Russian sanctions expert, Ivan Timofeev (2021), asserted that previous sanctions following Russian actions in Georgia 2008, Ukraine since 2014, and Syria since 2015, were “insufficient” and that the Kremlin hence saw no “major checks on a new military campaign.”

Of course, aside from misplaced certainty and risk acceptance, other factors that we do not (yet?) have sufficient evidence of might also have played a role in the invasion and, hence, in the failure to predict it. To name just a few: We might yet underestimate the brute salience of Ukraine to the Russian regime (cf. McGlynn, 2023; Shevtsova, 2020). It is possible that Russian decisionmakers highly valued control over Ukraine, perceiving the issue to be ultimately indivisible, and any agreement with the West over Ukraine as so unreliable, that any chance at gaining more direct control over Ukraine seemed worth taking. This would be in line with Fearon’s (1995) arguments on issue indivisibilities and commitment problems as rational explanations of war. Elites in the Kremlin might have also expected more rally-around-the-flag effects from ordinary Russians (as suggested by Guriev, 2022; Hale, 2022). Psychological factors might have played a key role, too. Russian elites might have been under the impression that they were steadily “losing” Ukraine (cf. Polianskii, 2022), and, under a specific interpretation of prospect theory, inappropriately overvalued the aversion of that loss (cf. Driedger, 2023, p. 218).

Importantly, all the aforementioned factors can be modeled on utility-based frameworks by affecting risk acceptance or the expected risks and rewards of an invasion or alternative policies. However, we do not claim that utility-based models can capture all possible causal pathways to invasion. For example, all utility-based models are based on some assumptions of rationality and stipulate that actors generally follow a logic of consequences. Thus, the Russian decision might have been (partially) caused out of a sense of this being an “appropriate” rather than net-useful response to perceived Western or Ukrainian policy, a logic emphasized by theories of roles and norms in international politics. Furthermore, the decision might have also well been influenced by a logic of affect (Markwica, 2018).

Why were we wrong?

Prior to the 2022 invasion, evidence for these underappreciated factors of the Russian regime’s risk acceptance and tendency for misplaced certainty was scarce, especially when compared to evidence that the Russian regime had access to plenty of information that suggested high risks and costs of an invasion.

This imbalance of relevant evidence seems to partly stem from the fact that miscalculation and risk acceptance are rarely observable in as glaring a way as during the 2022 invasion, due to the obvious failure of the operation. While some previous actions of Russia could have been counted as tentative evidence for miscalculation and risk acceptance, they could also have reflected other underlying factors, such as plain calculated brazenness. Examples include Russian cyber-attacks on Estonia in 2007 (Driedger, 2021b), the various well-publicized assassinations (e.g., of Sergei Skripal and Zelimkhan Khangoshvili) that were most likely carried out by Russian secret service members in Western countries, the hacking of the German parliament in 2015, and Russian interference in the 2016 US presidential elections. Similarly, some have previously warned of the Russian intelligence service’s tendencies to act on their own and to play to biases in the Kremlin (e.g., Galeotti, 2016), but this alone would have arguably been insufficient for confidently predicting miscalculations of the magnitude as in early 2022.

Another reason for the mispredictions seems to lay in prominent findings on long-term trends in Russian policy and how they were used to infer causal pathways when making predictions, leading analysts to focus on high costs and policy alternatives, rather than on miscalculation, risk acceptance, or other factors. Three bodies of research are relevant here. First, various studies have established that Russian politics are strongly shaped by elite corruption and rent-seeking (e.g., Adomeit, 2022; Yakovlev, 2021). Descriptions of Russia as a “kleptocracy” (Dawisha, 2014) or “mafia state” (Harding, 2021) have gained wide prominence among scholars and broader audiences.

Second, pre-invasion research into Russian coercive statecraft has found that, in its military operations, Russia has displayed some attention to reducing risks of international backlash, domestic dissatisfaction, and conflict entanglement (e.g., Dembinski & Polianskii, 2021; Kofman & Rojansky, 2018; Sutyagin, 2018). Rather than brute force, Russia often used sub-kinetic “hybrid,” and information measures to avoid the costs and risks of war (e.g., Fridman, 2018). This ties into broader lines of research on how, in modern times, war rarely pays off for attackers (e.g., Van Evera, 1999). However, there had not been a pre-2022 study to measure trajectories of Russian risk sensitivity in offensive military operations over time (for a post-2022 study, see Driedger, 2023). Thus, with evidence of some Russian risk aversion in the past, some observers were probably biased to presume more risk aversion than turned out to be warranted.

Third, a robust research program has identified regime survival as a key driving force in all areas of Russian politics (Frye, 2021; Greene & Robertson, 2019), be it economic policy (Miller, 2018), parliamentary politics (Reuter, 2017), or force structure and military doctrine (Renz, 2019).

To analysts seeking to forecast future Russian actions before the invasion, these three bodies of well-grounded research suggested causal pathways to non-invasion, as elites should, in most contexts, avoid large-scale warfare to maintain elite profits, avoid the risks of military entanglement, and ultimately ensure regime security. As the various caveats and scope conditions of these studies are difficult to apply to future cases in different contexts, analysts probably tended to focus on the more tangible aspect of expectable costs.

We find support for this view in some forecasts that expressly downgraded the potential causal role of misplaced certainty and risk acceptance. For example, applying apparent lessons from past conflicts to the prospects of invasion in early 2022, Yilmaz (2022) stated that “the Russian government has had a clear understanding of the risks on the ground. It has made a careful cost-benefit analysis and established clear and limited goals for the use of hard power.” Yilmaz further stated that this “cost-effective policy is a conscious choice because the Russian decision-makers know well that they do not have the means to maintain a large-scale war.” Some of these propositions, like those on a “clear understanding” or a “careful cost-benefit analysis” may be consistent with past data, but are not robustly corroborated by them.

Similarly, Chausovsky (2021) wrote that, when resorting to military force, the Russian regime has proven to be “rather conservative and risk-averse, with a strong cost-benefit analysis.” Ullman (2022) stated that “Putin also knows that an armed attack or aggressive use of force will make any chance of his achieving [his goals] even less likely than landing an astronaut on the sun. He also knows that the costs to Russia and to him personally will be high and possibly unaffordable.”

However, numerous other accounts explicitly acknowledged that their forecasts were only probabilistic, that unknown factors were likely at play, and that certain underlying assumptions, while plausible, were not robustly corroborated. For example, while making the cost-based case that Russia would likely not invade, Guriev (2022) pointed out that expected benefits unknown to us might exceed the expected costs of invasion. Surveying likely domestic repercussions, Hale (2022) cautiously inferred that the invasion might not happen, but only if “Putin understands all this, as one hopes he does.” Hunter (2022) surmised similar caveats, as did Timofeev (2021, 2022). Kagan et al. (2021) advised to “look for solid evidence that Putin’s thought process and calculations really have changed so fundamentally that he would either overlook these problems or accept these costs before accepting at face value the invasion plan he is ostensibly pursuing.” Vacroux (2022b) stated that key Russian elites might have convinced themselves that Ukrainians would abandon the central government in case of an attack.

We believe that an analysis of the forecasts itself does not allow any robust inferences on why some scholars were more aware of the limits of their projections while others were more confident in them. However, it stands to reason that cognitive shortcuts—usually called heuristics and biases—go a long way in explaining this difference. In particular, a long research program in psychology has established that “availability heuristics” (e.g., Tversky & Kahneman, 1973) tend to cause people ascertaining unfamiliar situations to put unwarranted confidence in the epistemic utility of data, trends, and causal mechanisms that they can easily recall. Our findings tacitly support this view, as the forecasts that fared better tended to reflect explicitly on the known and unknown presumptions of their projections, and were more guarded in their projected likelihood, actively working against the biasing effects of availability heuristics.

Relatedly, we find that the more successful forecasts based their predictions on a multitude of probably applicable causal factors while seeking to be mindful about clear definitions, scope conditions, potential, unknowns, and interactions. In line with Tetlock’s (2017) broader findings on political forecasting, this style of forecasting turned out to be more successful than forecasts relying on fewer causal mechanisms and factors whose validity for the Russian case was often more alleged rather than actively corroborated. In the language that Tetlock (2017) borrowed from Isaiah Berlin (1953/2013), foxes outperformed hedgehogs.

Conclusion: Lessons learned

Our findings corroborate existing insights into geopolitical forecasting: Rare and consequential political events, like war onsets and military escalation, are hard to predict, due to the multitude of relevant factors, non-linear developments, interactions, and demonstration effects (cf. Gleditsch, 2022; Lustick, 2022).

Furthermore, the results corroborate that, in forecasting, deep knowledge about case-specific facts is as necessary as a firm grasp on applicable causal mechanisms and scope conditions that are grounded in explanatory theory and empirically corroborated (cf. Gleditsch, 2022; Lustick, 2022). Forecasters should thus move beyond the sometimes-invoked dichotomy between theory-based research and area studies. Rather, they should combine the two. Specifically, analysts should seek to explicitly distinguish between ontology, or what theory-based findings on causal pathways tell us on the workings of international politics and conflict (how A generally relates to B under the condition of C) and epistemology (what do we know about A, B, and C in the case at hand—and what do we not know?). Among the forecasts we investigated, those that were more mindful of these questions turned out more successful.

In practice, relevant data are often ambiguous or unavailable. Our analysis illustrates how available sources can strongly corroborate some factors affecting a situation but be scarce for others. Such an uneven distribution of relevant data can facilitate mistakes, such as taking absent evidence as evidence for absence, or projecting from corroborated factors alone, rather than taking possible interactions with other, possibly uncorroborated or unknown factors into account. As our analysis further highlights, forecasts become more precise when analysts actively work against potential availability heuristics by reflecting on the assumptions underlying their interpretations. Considering the poor performance of monocausal arguments in the forecasts discussed here, particular attention should be paid to potential interactions and scope conditions of causal factors, requiring explicit and eclectic theorization (Sil & Katzenstein, 2010).

With all this in mind, forecasters have a heightened responsibility to both insist on established truths, even if they are uncomfortable to the target audience, while also practicing and promoting Socratic humility: we should try to get to know how much we do not know, as filling these gaps with heuristics and misplaced certainty will yield flawed and potentially disastrous predictions and recommendations.

Our findings suggest some direct roads for future research: Our analysis of the most prominent English-language accounts could be broadened into a more systematic survey of a larger sample size to further examine how and why predictions and underlying arguments differed from each other. Of particular interest would be if the professional background or location of the forecasters systematically affects prediction outcomes and argument structures. Related (mis-)predictions should also be mapped and investigated, like those on the performance of the Russian military, which most in the West (and Russia) seem to have gotten wrong.

Acknowledgements

We would like to thank Jule vom Köhlerwald as well as the editors and anonymous reviewers at Contemporary Security Policy for their valuable feedback.

Disclosure statement

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

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Additional information

Funding

Part of this research was made possible within the framework of the research project “Regional Research Center Transformations of Political Violence” [01UG2203A], funded by Germany’s Federal Ministry of Education and Research.

Notes on contributors

Jonas J. Driedger

Jonas J. Driedger is a Researcher at the International Security Department of the Peace Research Institute Frankfurt (PRIF) and at the Regional Research Center Transformations of Political Violence (TraCe). He specializes in the causes of military conflict as well as foreign and security policy with a focus on Germany, Russia, and transatlantic relations. He has conducted extensive fieldwork and interviews in Ukraine and Russia. Previously, he was DAAD Fellow at the Henry A. Kissinger Center for Global Affairs, School of Advanced International Studies (SAIS) in Washington, DC, Visiting Researcher at the Higher School of Economics in Moscow (HSE), AICGS/DAAD Resident Fellow in Washington, DC, and Officer for Security Policy at Haus Rissen in Hamburg.

Mikhail Polianskii

Mikhail Polianskii is a Junior Research Associate at the Peace Research Institute Frankfurt (PRIF) and a PhD Candidate at the Goethe University Frankfurt. His research focuses on Russian foreign policy and the crisis of European security order in the post-Cold War era. Born in Russia, he graduated from MGIMO (Moscow State Institute of International Relations) in the course of which he interned at the Russian Foreign Ministry (2018, 2019). He conducted fieldwork and interviews in the framework of his PhD studies as a visiting researcher at the Center for Comprehensive European and International Studies (CCEIS) at the Higher School of Economics in Moscow (HSE). He has published several research articles on these topics and taught an advanced graduate course on Russian foreign policy.

Notes

1 Russia had engaged in military and “hybrid” operations on Ukrainian territory since at least 2014. For brevity’s sake, we henceforth use “invasion” to denote the large-scale Russian operation into hitherto unoccupied Ukrainian territory since February 24, 2022.

2 We take “forecasting” as a “process of generating falsifiable predictions about future geopolitical events” that involves formal (percentages) or informal (e.g., “more likely than not”) expressions of probability ranges as well as the scoring and evaluation of such predictions (cf. Horowitz et al., 2021).

3 With such terms, we simply mean that a result that was deemed unlikely did, in fact, occur (cf. Lustick, 2022).

4 For such limited concepts of instrumental rationality, and the more demanding counterpart of procedural rationality, see Zagare (1990).