Abstract
Research on government evaluations after disasters focuses on democracies and retrospective decision-making theory, where voters either attentively evaluate performance based on experiences or make blind decisions reflecting emotional dispositions. This study examines individuals’ appraisals of government performance across three dimensions following the 2023 earthquakes in Türkiye: disaster management, preparedness, and blame attribution. Using nationally representative survey data, we find that earthquake losses and negative emotions relate to negative evaluations of government preparedness and management. However, these factors show no association with blaming the government, suggesting voters distinguish between different dimensions of government responsibility. Opposition party support and social media trust show significant negative associations across all dimensions, with partisanship being the strongest determinant. These findings highlight the relationship between affective polarisation and divergent government evaluations, suggesting that in polarised contexts with contrasting media, affective polarisation may be a stronger predictor of disaster-related political outcomes than retrospection theory.
Keywords
Introduction
The improved documentation of an increasing number of extreme weather events and disasters worldwide highlights their economic, social, and political significance (Alimonti and Mariani, 2024; Tin et al., 2024). Academic interest is also rising in how disasters relate to the public’s political perceptions and attitudes. However, gaps remain in our understanding of how social, political, and individual factors are associated with government evaluations after catastrophic events. Most political science literature uses the retrospective decision-making framework and focuses on US politics and democratic countries (Demirdogen and Olhan, 2025: 330–332; Rubin, 2020: 244). Thus, research findings are valid mostly in contexts where free media exists, allowing citizens to make more informed political decisions after disasters.
We contribute to this expanding literature by analysing the February 2023 earthquakes in Türkiye, which claimed more than 53,000 lives and resulted in the loss of around 9% of the country’s gross national income (Kahramanmaraş ve Hatay Depremleri Yeniden İmar ve Geliştirme Raporu, 2026). We utilise available questions in a nationally representative face-to-face survey conducted approximately 2 months after the disaster to assess the relationship between evaluations of government performance and social and political factors. At the time of the earthquakes, Türkiye was classified as an electoral autocracy (Nord et al., 2024). What makes such a non-democratic country interesting is that the government controls narratives through mainstream media but cannot fully suppress all dissent. Turkish politics is also highly and affectively polarised (Lauka et al., 2018), with relatively vibrant opposition parties and civil society activity. Despite the threat of censorship and suppression, this activism results in the circulation of alternative narratives through opposition parties and social media.
In this non-democratic and polarised context, the 2023 earthquakes generated starkly different narratives between the government and opposition parties. While the opposition blamed the government for inadequate response, corruption, and regulatory failures, the governing coalition framed the disaster as an unavoidable ‘disaster of the century’ and emphasised successful rescue efforts through pro-government media. These divergent interpretations, reinforced by different media sources, meant that government and opposition supporters were exposed to fundamentally different framings of both the disaster’s causes and the state’s response.
We analyse how the Turkish public evaluated government performance after the earthquakes within this context of electoral autocracy and polarisation. Similar to the ‘traditional disaster management cycle’ (Sawalha, 2020: 471), we use available survey questions to distinguish how the public evaluated government performance across different dimensions of government responsibility: the pre-disaster phase of preparedness, the overall disaster management, and blame attribution for the losses.
We first develop hypotheses from retrospective decision-making theory (Healy and Malhotra, 2013) to examine whether political appraisals were associated with actual experiences of loss, as attentive retrospection would predict (Arceneaux and Stein, 2006). Then we analyse if general emotional dispositions were associated with perceptions of government performance, as blind retrospection theory would expect (Achen and Bartels, 2017). The latter focus on negative emotions is also warranted, given that the Turkish earthquakes were a devastating event that could be categorised as collective trauma. As such, the scholarly expectation would be for the traumatic event to have political implications (Marsh, 2023) and for the subsequently and widely felt negative emotions to be linked to government evaluations (Atkeson and Maestas, 2012).
While retrospective decision-making theory emphasises the role of experiences and emotions in shaping political evaluations, pre-existing political divisions might be an equally or more significant factor in perceptions. In affectively polarised contexts, party support is associated with government evaluations after natural disasters (Malhotra and Kuo, 2008). Given that Türkiye was a highly polarised country at the time of the earthquakes, we complement our analysis of retrospection theory with an examination of party support.
Our findings regarding earthquake-related experiences, emotions, and perceptions of government performance corroborate both blind and attentive retrospective decision-making arguments. In other words, both the experience of loss and negative emotions were related to negative evaluations of the government’s disaster management and preparedness, independent of each other and other factors. Notably, however, these factors showed no association with blaming the government, suggesting that citizens distinguished between government performance failures and the cause of the losses.
Furthermore, our analysis clearly shows that, as expected, individuals had different reactions after the earthquakes based on partisanship. Specifically, opposition party support and trust in social media increased the likelihood of negative government evaluations. Crucially, the magnitude of partisan support was higher than any other factor. Based on these findings, we urge academic studies to focus more on different political contexts when analysing the relationship between natural disasters and government perceptions. In an autocratic and affectively polarised context, political framing can make pre-existing political identities more prominent than other social factors, unlike in democratic settings.
This finding further contributes to two closely related literatures on natural disasters: one on electoral outcomes and the other on regime change. Regarding the first literature, retrospection theory speaks directly of voting behaviour although there are no consistent findings. Attentive retrospection predicts that voters punish poorly performing incumbents in the elections (e.g. Eriksson, 2016). Conversely, they reward politicians who respond to the disaster effectively, often through increased relief spending (Bechtel and Hainmueller, 2011; Fukumoto and Kikuta, 2024; Healy and Malhotra, 2009). Blind retrospection expects voters to punish incumbents in elections irrespective of responsibility, while research on affective polarisation argues that electoral outcomes reflect partisan loyalties rather than performance. Although much of this literature focuses on advanced economies, findings from developing countries are similarly mixed. Some studies report increased support for the government or reduced anti‑incumbent voting (Baerlocher et al., 2025; Gallego, 2018), while others show shifts in electoral preferences and evidence of voters punishing governments at the polls (Cole et al., 2012; Visconti, 2022). Taken together, this body of work suggests that although disasters can influence election outcomes, neither the occurrence nor the direction of such effects is guaranteed. Instead, their impact depends heavily on the broader political context, such as the type of regime and state capacity, and on contingent factors like the location and timing of the shock relative to the election (for a review, see Hosseinkhani, 2025).
Turning to a related literature, large-N datasets covering global cases of natural disasters across countries and several decades suggest that earthquakes and other hazards can create openings for democratisation – if not through immediate regime breakdown, then at least through short-term improvements in a number of democratic indicators (Povitkina et al., 2025; Rahman et al., 2017). A growing literature also shows that institutional failures, such as corruption, weak public administration, and inadequate planning, significantly increase death tolls and material losses in the aftermath of natural disasters (Kahn, 2005; Persson and Povitkina, 2017; Sjöstedt and Povitkina, 2017). Consequently, severe natural disasters are expected to reveal the detrimental effects of authoritarianism and poor institutional quality, prompting citizens to respond through civil society mobilisation or electoral channels.
Türkiye’s political history suggests similar patterns. The coalition government’s poor performance following the 1999 Marmara earthquakes, which claimed around 18,000 lives, catalysed civil society mobilisation and triggered a dramatic electoral realignment (Kubicek, 2002). Occurring in a much less polarised and more democratic political context, widespread perceptions of corruption and administrative incompetence resulted in all three governing parties failing to re-enter parliament (Akarca and Tansel, 2016). In their place, the newly established Justice and Development Party (Adalet ve Kalkınma Partisi, AKP), led by Recep Tayyip Erdoğan, secured a decisive victory in the 2002 elections, ushering in a new political era, which initially brought about further democratisation.
Ahead of the 14–28 May 2023 presidential and parliamentary elections, Türkiye faced similar circumstances. As the following sections will show, there was again widespread belief in government ineffectiveness managing the crisis, contributing to higher levels of destruction and loss of life. The country was not only classified as an electoral autocracy but also scored 36 out of 100 in the 2022 Corruption Perceptions Index, below the global average, and ranked 101st out of 180 countries (Corruption Perceptions Index, 2022). Against this backdrop, many, including the opposition, expected a similar pattern as 1999 and anticipated the governing coalition to lose the elections (Yardımcı-Geyikçi and Yavuzyilmaz, 2024: 269). Given Türkiye’s political regime at the time, such a change could have also signalled a process of democratisation, consistent with global findings that natural disasters can at times catalyse political opening and institutional change.
Our analysis offers a refined explanation for why these expectations did not materialise and why there was no political change after the February earthquakes. Our survey includes a question on voting intentions but lacks data on actual voting behaviour, making it impossible for us to directly assess the electoral consequences of earthquakes for individual voting. However, existing research on the May 2023 elections has already produced informative findings, demonstrating that the earthquakes did not affect support for the governing coalition even in the most impacted provinces (Demirdogen and Olhan, 2025; Esen et al., 2023). Moreover, there is considerable evidence to suggest that the AKP maintained its voter base partly because of partisanship (Aytaç et al., 2025; Çarkoğlu, 2024). Our results contribute to this discussion by suggesting that partisan bias and affective polarisation overshadowed negative evaluations of the government’s earthquake performance, despite the scale of human and material losses, contributing to the election results. In other words, government supporters interpreted the government performance regarding the earthquakes in a way that aligned with their partisan loyalties, incorporating it only marginally, if at all, into their voting decisions.
The rest of the article is organised as follows. First, we review the literature that informed our hypotheses on the relationship between disaster-related loss, emotions, and political evaluations. After we provide background information on the Turkish earthquakes of 2023, in the subsequent two sections, we explain our methodology and share our results. Finally, we discuss our findings before we conclude.
Disasters, emotions, and government evaluations
A growing body of literature examines the relationship between exogenous events, such as disasters and politics. Despite earlier evidence of the minimal effect of disasters on political outcomes (Abney and Hill, 1966), the current literature strongly suggests that these types of events are associated with the electorates’ evaluations of the incumbent’s performance and policy preferences. However, the direction of the association is not consistent, and most of the literature focuses either on the United States or other Western countries (for exceptions, see, for example, Cole et al., 2012; Demirdogen and Olhan, 2025; Fair et al., 2017; Gallego, 2018). While retrospective voting serves as the primary theoretical framework, almost all research on the matter assumes the presence of a democratic regime, where citizens can freely make political decisions and assessments (Rubin, 2020).
According to retrospective decision-making theory, citizens form political opinions about the incumbent based on their experiences during the tenure of the government. Simply put, the theory posits that ‘voters reward incumbent officials for good times and punish them for bad ones’ (Huber et al., 2012: 720). Within this general theory, the blind retrospection thesis argues that voters punish incumbents for shocking and devastating events, such as shark attacks, floods, or droughts, even when these events are beyond the control of governments (Achen and Bartels, 2017; Heersink et al., 2017). By contrast, the attentive retrospection hypothesis suggests that the electorate is more ‘rational’ and might blame or reward incumbents based on their actual performance and actions, especially in relief efforts after a disaster (Ashworth et al., 2018; Bechtel and Hainmueller, 2011; Cole et al., 2012; Gasper and Reeves, 2011; Healy and Malhotra, 2009). Besides voting, individuals can also change their political attitudes, preferences, or behaviour after natural catastrophes by, for example, favouring action or policies on climate change (Arias and Blair, 2024; Garside and Zhai, 2022).
In the Turkish case, the retrospection thesis in general would lead us to expect that those who personally suffered ‘bad times’ because of the earthquakes would punish the incumbents (Arceneaux and Stein, 2006). As will be explained below, the number of lives lost and the damages incurred after the 2023 earthquakes were unprecedented. There were also accusations from opposition groups that the government was responsible for the scale of the tragedy. For this reason, our first hypothesis regarding evaluations of government performance is in line with attentive decision-making:
Hypothesis 1. Individuals who suffered disaster-related losses were less likely to evaluate the government’s performance positively.
As an alternative to attentive decision-making, political responses after disasters can be considered from the perspective of emotions. In general, there is considerable evidence to warrant a deeper look at negative emotions, such as anger, fear, and sadness, and how they are associated with political decisions (e.g. Marcus, 2000; Valentino et al., 2008; Ward et al., 2024; Weber, 2013). Research on the relationship between natural disasters and politics also shows that emotions, such as anxiety and anger, are associated with negative perceptions of the government (Atkeson and Maestas, 2012).
Such a relationship between negative emotions and government evaluations after a natural disaster exists in part because catastrophes, such as hurricanes, floods, tsunamis, or earthquakes, are a type of traumatic event that overwhelms individuals, making them feel vulnerable, powerless, and in shock. Traumatic events involve death, threats to life, or intense aggression (Marsh, 2023: 1038) and can affect both individuals and groups (Vertzberger, 1997: 864). While man-made collective traumas such as terrorist attacks (Hall and Ross, 2015; Huddy and Feldman, 2011), mass shootings (García-Montoya et al., 2022), arson attacks (Marsh, 2023), and political assassinations (Das et al., 2024; Vertzberger, 1997) are related to intense feelings like pain, grief, and anger, natural disasters correlate with similar emotions, including post-traumatic stress, depression, and anxiety (Goenjian et al., 2000; Neria et al., 2008; Wang et al., 2020).
Within the broad category of catastrophic events, earthquakes lead to higher levels of post-traumatic symptoms due to their unpredictability (Wang et al., 2019: 1071). Deadly earthquakes claim lives in gruesome ways, trap individuals under collapsed buildings, cause loss of property and homes, displace communities, and result in numerous physical and emotional injuries. Reported rates of post-traumatic stress disorder (PTSD) among earthquake survivors globally vary from 12% to 87%, indicating higher levels of diagnosis compared with other disasters. Lifetime PTSD is also common, with estimates of around 8%, alongside other psychiatric conditions like generalised anxiety disorder or phobias (Farooqui et al., 2017: 138).
While negative emotions may be related to the disaster experience itself and therefore point to attentiveness, the blind decision-making theory implies that moods might play an independent role, unrelated to the disaster experience. To put it differently, ‘in addition to predictable cognitive biases, emotions may influence voters so that they sometimes reward or punish politicians simply because they feel happy or sad for reasons that have nothing to do with incumbent performance’ (Healy and Malhotra, 2013: 289; also see Liberini et al., 2017). Therefore, our second hypothesis links emotions to evaluations of incumbent performance both as an independent factor and in interaction with earthquake-related loss as follows:
Hypothesis 2a. Individuals who felt negative emotions were less likely to evaluate the government’s performance positively.
Hypothesis 2b. Individuals who suffered disaster-related losses and felt negative emotions were less likely to evaluate the government’s performance positively.
Beyond individual loss and emotions in the aftermath of disasters, research in settings with high levels of affective polarisation, such as the United States, suggests that political evaluations in the aftermath of exogenous events are associated with party support. On one hand, there is evidence to support attentive decision-making, where voters evaluate the government based on how it actually performs (Arias and Blair, 2024). For example, in a comprehensive analysis of perceptions of government performance after Hurricane Katrina, Atkeson and Maestas (2012) demonstrate how even some of the Republican voters changed their perceptions and held President Bush accountable. This was because a ‘focusing event’ in the form of a natural disaster (Birkland, 1997, 1998) led to greater media coverage, which was followed more attentively by an anxious public, and resulted in the re-evaluation of policies.
On the other hand, there is evidence that partisanship and previous social identities become more salient because of exogenous events, suggesting that they introduce bias to what could otherwise be attentive decisions. Individual-level interpretations of traumatic events depend on the social identities of both the victims and witnesses (Muldoon et al., 2021). Positionality can leave subordinate groups in more vulnerable positions due to actual and perceived powerlessness within social and political structures. By contrast, group identities can be reaffirmed and strengthened if the community reorganises and expands through collective mobilisation towards various goals, such as aid and rescue, solidarity promotion, social justice, and political action (Drury et al., 2009; Jenkins, 2019). Thus, how individuals experience disasters does not happen in a social vacuum but is often shaped by pre-existing group identities.
Political party support in highly polarised settings can serve as one form of social identity because ‘inter-party ill will is sufficient to inject partisanship into decisions that are entirely personal’ (Iyengar et al., 2012: 418). Political leaders also play a central role in framing the meaning of the event and guiding public interpretation. As the cultural strand of trauma literature suggests, ‘events are not inherently traumatic’ (Alexander, 2004: 8) and traumas are ‘historically made, not born’ (Smelser, 2004: 37). They are constructed through a sociocultural process involving human agency or ‘carrier groups’, who make claims and create master narratives, which identify the pain, victims, perpetrators, and the significance of the events for the wider group (Alexander, 2004). These carrier groups can be a range of actors, such as religious leaders, intellectuals, journalists, and politicians, who tap into negative emotions that echo the victims’ experiences (Lerner, 2022: 42) and frame a disaster as traumatic (Smelser, 2004: 41).
In any given context, these affective narratives are usually contested, and the impact of trauma is dependent on the resources and power of the carrier groups, as well as the different frames available to victims and witnesses. In contexts of preset affective polarisation, political actors already generate controversial and opposing ‘truths’ on a wide variety of issues. As a result, ‘people increasingly perceive and describe’ events from a hostile lens that divides politics and society into groups of ‘us’ and ‘them’ (Somer and McCoy, 2019: 13). In the aftermath of a catastrophic event, such mass affective polarisation would be linked to continued support for the government for those who already pay allegiance to the incumbents. This is why examining Hurricane Katrina’s political consequences, Malhotra and Kuo (2008: 127) argued that although it was possible for party allegiance to be mitigated by other relevant information, ‘Republicans and Democrats observed the same events through different lenses, underscoring the importance of party attachments in explaining reactions’. Based on similar expectations, our third hypothesis takes party support into account independently and in interaction with negative emotions, as follows:
Hypothesis 3a. Individuals who supported opposition parties were less likely to evaluate the government’s performance positively.
Hypothesis 3b. Individuals who supported opposition parties and have negative emotions were less likely to evaluate the government’s performance positively.
An additional factor to consider is the role of the media in framing events and government performance. It is well established that the media sets the agenda, determines the most salient issues, and creates attention beyond those who live in geographical proximity to the disaster. Higher media presence because of a disaster can favour incumbents before elections (Masiero and Santarossa, 2021). However, the media can report the government’s response positively or negatively (Albrecht, 2022: 19), or can depict events through the heroic efforts of state rescuers (Venugopal and Yasir, 2017). Thus, the media not only create spectators but also direct what citizens should be feeling and how they should be evaluating government performance (Albrecht, 2022; Atkeson and Maestas, 2012; Barnes et al., 2008).
Because in Türkiye traditional media is controlled by the government and they propagate a more positive framing of the state response after the earthquakes, we would expect fewer negative reactions among those who trust these media outlets (Wu et al., 2021). By contrast, those who trust social media have a higher likelihood of believing negative portrayals of government performance and greater exposure to ‘videos, pictures, and stories of trauma’ (Marsh, 2023: 1039), which increases their chances of experiencing negative emotions. Trust in the source of media is also likely to work in tandem with party identity (Iyengar et al., 2012), as those who are supportive of the incumbent are more likely to follow traditional media, while those who favour the opposition are likely to seek and trust alternative information from social media platforms. In other words, we expect that in Türkiye, due to polarisation and electoral autocracy, different political groups trust different kinds of media and were exposed to different narratives of the earthquakes. As a result, we expect trust in social media would independently relate to government evaluations and its effect would increase in interaction with emotions, as follows:
Hypothesis 4a. Individuals who trusted social media for disaster-related news were less likely to evaluate the government’s performance positively.
Hypothesis 4b. Individuals who trusted social media for disaster-related news and felt negative emotions were less likely to evaluate the government’s performance positively.
In summary, the literature suggests that government evaluations following natural disasters are associated with multiple factors, potentially interacting with each other. While retrospective voting theory provides the foundational framework, empirical evidence points to three key mechanisms through which disasters relate to political assessments:
Direct personal experiences of loss independently relate to negative government evaluations, along with the reinforcing effect of negative emotion on individuals who have experienced losses (attentive retrospection).
Negative emotions operating independently of actual government performance relate to negative government evaluations (blind retrospection).
Pre-existing political loyalties, which are linked to different interpretations of events and emotions through political leadership and media sources, relate to government evaluations (affective polarisation).
February 2023 earthquakes and the political context in Türkiye
On 6 February 2023, 11 southeastern cities of Türkiye were struck by 7.8 and 7.7 magnitude earthquakes, occurring 9 hours apart. Although Türkiye is situated on geological fault lines and is no stranger to tremors, the scale and tens of thousands of aftershocks made these the strongest earthquakes in modern Turkish history. The earthquakes directly impacted more than 13 million people in the 11 cities, approximately 16% of the Turkish population. More than 53,000 people died, and 107,000 were injured. Around 35,000 buildings collapsed, with hundreds of thousands more deemed damaged, including hospitals, schools, roads, and other key infrastructure. Within 3 months after the earthquakes, nearly 3 million people were living in tents and containers, with approximately the same number relocating to other cities (Abatay, 2024). Beyond physical disruption, the psychological toll was severe, with 51.4% of survivors experiencing PTSD symptoms 3 months later, higher than other earthquakes globally given the disaster’s unprecedented scale and widespread loss (İlhan et al., 2023).
These extraordinary levels of trauma and loss occurred against the backdrop of a divided political landscape. Before the earthquakes, Türkiye was already highly polarised between those who supported President Erdoğan and the AKP and those who supported opposition parties. More precisely, two main alliances were formed in advance of the May 2023 general elections and the biggest party in the opposition coalition National Alliance was the Republican People’s Party (Cumhuriyet Halk Partisi – CHP), followed by five other smaller parties.
Research on polarisation highlights political divisions in Türkiye. Global and comparative studies have found that the country exhibits the highest level of affective polarisation worldwide (Lauka et al., 2018; Reiljan et al., 2024). Other studies have shown that partisan social identities are associated with negative feelings, perceptions of threat, social distance, and intolerance towards other party supporters (Erdoğan, 2018; Laebens and Öztürk, 2021). Emotional polarisation is also evident with different party supporters feeling anxious about diverse future scenarios (Erdoğan and Uyan Semerci, 2018).
Affective polarisation in Türkiye is also highly associated with the media. The increasingly authoritarian AKP government controls 90% of broadcast media and represses the rest (Andı et al., 2025: 6), which creates unfair conditions before the elections. Due to repression, opposition broadcast media also sometimes refrain from overly criticising the government, fearful of censorship or prosecutions. After the 2023 earthquakes, for example, opposition newspapers largely avoided adopting a critical stance towards the government’s handling of the earthquakes (Binici et al., 2025). However, newspaper readership is relatively low, and television is the predominant source of news for a significant majority of citizens (Erdoğan and Uyan Semerci, 2018).
While a higher percentage of government supporters rely on television for news (61%), Instagram and WhatsApp are common sources of news among opposition voters (Andı et al., 2025: 12). There are also active opposition YouTube news channels that reach tens of thousands of viewers every day, operating similarly to broadcast TV, with correspondents throughout the country, regular bulletins, breaking news reporting, and frequent commentaries. Türkiye is one of the top 10 countries in YouTube audience size in the world, with a 76% penetration rate, while it is also the most preferred source of news among social media platforms (Dierks, 2024). Similar to television, there are alternative sources of information on social media, including YouTube. However, usage does not mitigate polarisation and may even exacerbate it through the creation of echo chambers (Aytaç et al., 2025).
Given the Turkish context of affective polarisation and incumbent media control, the narratives of the 2023 earthquakes differed between the AKP and the opposition. The National Alliance and opposition civil society groups held the government responsible for the calamity, using social media platforms, including X/Twitter (Görkemli et al., 2024). Criticism included the slow rescue and aid response due to the ‘over-centralisation of [the] Turkish emergency management system’ following Türkiye’s transition to presidentialism in 2017 (Ertas, 2024: 628-629). Nepotism was highlighted since the disaster and emergency authority was headed by officials without the necessary expertise. Widespread corruption was also blamed, as the government allowed favoured construction companies to build deficient infrastructure (Aksoy and Çevik, 2023). Since 2018, contractors and dwellings have been given amnesties in exchange for a fee, resulting in half of all developments in Türkiye not complying with regulations (Cifuentes-Faura, 2024: 3).
For its part, the governing coalition framed the earthquakes as the ‘disaster of the century’, suggesting that no administration could have prevented the catastrophe, and accused private construction companies, building owners, and opposition parties (Goldring et al., 2025). It compared its efforts with the 1999 earthquake, highlighted its achievements, and promised to quickly rebuild homes. Pro-government media focused on heroic and successful stories of rescue and aid, while the government repressed alternative narratives by an X/Twitter ban (later repealed), threatening dissenting voices and calling critiques ‘traitors’ (Eissenstat, 2023). Overall, ‘the government’s crisis management strategy was based on explaining the action plan and showing its performance via visual and numerical data’ (Şen, 2023: 75).
Data, variables, and methodology
The data for this article are based on an original public opinion survey conducted by KONDA Research and Consultancy Inc. on 29–30 April 2023, as part of a larger project. The survey included 2811 face-to-face interviews across 12 regions, 36 provinces and 110 districts, with one province from each of Türkiye’s 26 statistically representative sub-regions according to the Turkish Statistical Institute. Age and gender quotas were applied to a minimum of 18 face-to-face surveys conducted in 150 neighbourhoods and villages, ensuring equal representation based on these demographic categories.
The analysis for this article has one dependent variable, namely evaluations of government performance regarding the earthquakes. We used available questions related to the earthquakes in the survey 1 to capture different dimensions of government performance. The traditional disaster management cycle framework distinguishes between pre-disaster, disaster, and post-disaster phases of response (Coetzee and Van Niekerk, 2012; Sawalha, 2020). Similar to this, we created three subcategories for the dependent variable, measuring different aspects of incumbent responsibility in overall disaster management, preparedness, and blame attribution for the losses. To analyse the respondents’ evaluations of disaster management, we utilised the question, ‘When you think about the earthquake disaster that happened two months ago, how would you rate the state’s disaster management?’ We generated an ordinal variable with five categories using the response rates ranging from 1 (Very unsuccessful) to 10 (Very successful). To measure perceptions on preparedness, we used the question ‘Were the government’s preparations adequate before the earthquake occurred?’ (1 = Yes, adequate, 0 = No, inadequate). Finally, to decipher blame attribution, we used answers to the question ‘Who do you think is most responsible for the losses in the earthquakes?’ (1 = Government/Ministries, 0 = local governments, contractors, building owners, foreign powers, or ‘no one is responsible as this is a natural disaster’).
As explained in the above section, our analysis has four independent variables: earthquake-related losses, negative emotions, party support, and trust in media sources. We created variables based on the available questions. Earthquake-related losses were measured with the survey question ‘Did you lose any of the following due to the earthquakes last February?’ (1 = Loss of a family member, relative, acquaintance or house, 0 = No loss). Regarding feeling negative emotions, we used the survey question ‘how often have you felt any of the following in the last week?’ that separately listed anxiety, sadness, and anger with the response categories of never, sometimes, often, and always. This question asked about general moods and did not specifically prompt the respondent to consider the earthquakes, making it ideal to test blind retrospection. As such, we created a dichotomous variable to capture the overall disposition of the respondents and whether they felt any of the three negative emotions in the last week (1 = Often and always, 0 = Otherwise).
To understand party support, respondents were asked, ‘If parliamentary elections were held today, which party would you vote for?’ This was an open-ended question, and we grouped party support into two categories based on those who declared that they would vote for one of the parties not included in the governing coalition (1 = Opposition, 0 = Governing coalition). Finally, we measured the independent variable of media source with the question ‘Which news source did you trust the most regarding the earthquakes?’ by generating a binary variable (1 = Social media, 0 = Otherwise).
To test the four main hypotheses and their interaction with emotions in sub-hypotheses, we ran multivariate ordinal and logistic regression analyses on the survey data. Control variables included gender (1 = Female, 0 = Otherwise), the distance of the survey locations to the earthquakes’ epicentre, Kahramanmaraş, based on the Ministry of Transportation and Infrastructure Intercity Distance Table (Karayolları Genel Müdürlüğü, 2026) (continuous and scaled at 100 km, min = 0.78 and max = 12.32), age (ordinal discrete variable, 1 = 18–32, 2 = 33–48, 3 = 49+), education (1 = College graduates, 0 = Otherwise), ethnicity (1 = Turkish, 0 = Otherwise), and income (continuous and scaled at 2000 TL, min = 1 and max = 55). All dependent variables were binary, except disaster management, which had five categories (1–5). Therefore, we conducted logistic regression analysis for government preparedness and blame attribution, whereas disaster management was analysed through ordinal logistic regression.
Results
This section presents the results from our regression analysis based on the public opinion survey and variables described in the methodology section. 2 Table 1 shows factors associated with perceptions of disaster management, preparedness, and blame attribution. 3
Ordinal logistic and logistic regression analysis of government performance.
Exponentiated coefficients; Standard errors in parentheses *p < 0.10, **p < 0.05, ***p < 0.01.
Control variables showed no consistent patterns. In the preparedness model, female 4 and college-educated respondents were more critical of the government, whereas older respondents more positive. College-educated and older respondents attributed more blame to the government, while ethnically Turkish and older respondents evaluated government’s management of the disaster more positively. The location of the interview, measured through distance from the epicentre of the earthquakes, was insignificant except for blame attribution: for every 100 km increase in distance from Kahramanmaraş, the odds of holding the government responsible for the losses increased by around 7%. 5
Regarding our independent variables, individuals who suffered earthquake-related losses were almost 50% less likely to view the government as prepared and 32% less likely to evaluate state-led disaster management positively. However, the experience of loss alone did not have a statistically significant relationship with blaming the government. These findings partially support Hypothesis 1 regarding the relationship between earthquake losses on perceptions of government performance. Our theoretical expectations in Hypothesis 2 were also partially supported. Having a negative emotional disposition decreased the likelihood of positive assessments of state-led disaster management as well as the government’s preparedness. 6 More precisely, individuals experiencing any negative emotions were one-third less likely to view the government as prepared for the disaster and rate the disaster management positively. Similar to loss, negative emotions did not independently predict blaming the government.
Partisanship produced the strongest and most consistent results across all models. In line with Hypothesis 3, supporting an opposition party was strongly connected to negative assessments of government performance across all models. 7 Opposition voters were considerably less likely to evaluate the government’s disaster management positively and rate the government as prepared for the earthquakes by roughly 98%–97% lower odds. They were nearly 12 times more likely to blame the government for the losses. 8 Furthermore, partisanship interacted with negative emotions in blame attribution. Opposition supporters who reported negative emotions were almost twice as likely as their counterparts, who did not experience negative emotions, to blame the government. This indicates that emotional reactions to the earthquakes reinforced partisan divides, amplifying critical attitudes among opposition supporters.
As expected from Hypothesis 4a, reliance on social media was also linked to more critical views of the government’s handling of the disaster. Individuals who trusted social media for earthquake-related news were 53% and 57% less likely to have positive perceptions of the government’s management and preparedness, respectively. They were also about twice as likely to blame the government. However, the interaction between social media reliance and emotions did not significantly relate to government evaluations across any of the models, contrary to Hypothesis 4b.
These results highlight two important findings. First, they demonstrate the necessity of distinguishing between different dimensions of government evaluations, and second, they confirm the pivotal role of political polarisation in shaping disaster accountability.
Discussion
Results from the April 2023 survey, as described above, partially support various retrospective decision-making arguments. There is supporting evidence for the attentive voter thesis (e.g. Gasper and Reeves, 2011), because our findings suggest that respondents evaluated the incumbent’s disaster management and preparedness based on their actual personal experiences (similar to Arceneaux and Stein, 2006). However, these personal experiences were not associated with blaming the government. This variation in results shows that citizens make a distinction between direct and indirect government responsibilities. In other words, when asked how the incumbent managed the disaster and whether it was prepared, respondents held the government responsible for their adverse personal experiences. However, given that this was a natural disaster and others were also responsible for damaged buildings, the incumbent’s role in lives and properties lost after the earthquakes was more diffuse and complicated, resulting in no significant association between personal experience and government responsibility. That the likelihood of blaming the incumbent rose with greater distance from the epicentre also suggests that those who physically felt the tremor may have viewed the disaster as more ‘natural’ than ‘man‑made’.
Although there is strong evidence to support attentive decision-making based on personal experiences, results also show some corroborating evidence for blind retrospection. Feeling any negative emotion – in other words, angry, anxious, or sad – was associated with some of the government performance evaluations. In line with blind retrospection (e.g. Achen and Bartels, 2017; Liberini et al., 2017), individuals who were in a bad mood were more likely to punish the incumbent, suggesting that negative affect served as a heuristic tool (Ward et al., 2024: 5). Although blame attribution is an outlier here as well, given that the survey question did not prompt the respondents to link their mood with the earthquakes, we cannot refute the independent effect of general emotions.
The retrospective decision-making literature focuses primarily on Western democracies, assuming free media that can adopt multiple frames independent of government pressure. Studies on the United States and other democratic contexts show that government evaluations depend on how well media informs the electorate, as media can counteract politicians’ attempts to spin narratives post-disaster (e.g. Atkeson and Maestas, 2012: 68; Healy and Malhotra, 2013: 300–301). Our Turkish survey data also highlight the significance of information sources in evaluating government performance. However, unlike fully democratic regimes, the incumbent controlled media and framed the dominant narrative after the earthquakes, portraying its performance positively. With access to independent news blocked except for a few TV channels and opposition newspapers refraining from criticism (Binici et al., 2025), social media became the primary source for independent disaster zone reporting. Consequently, trust in social media increased the likelihood of negative government performance evaluations across all models.
The results should also be interpreted within the broader debate on disaster governance and political accountability in non-democratic regimes (Kahn, 2005; Persson and Povitkina, 2017). Retrospective voting literature assumes institutional conditions (aside from free media, also fair political contestation and autonomous state agencies) more characteristic of established democracies than electoral autocracies. However, comparative work shows effective crisis and climate management does not map neatly onto regime type; authoritarian governments, especially if they are less corrupt and have higher institutional quality, can mobilise resources efficiently, centralise decision-making, and perform better, in ways that blur standard performance and accountability expectations (Carlitz and Povitkina, 2021; Persson and Povitkina, 2017; Povitkina, 2018). Furthermore, in presidential and majoritarian electoral systems, as well as in contexts with weak checks and balances or newer democracies, rent-seeking behaviour increases, leading to higher public spending after disasters to maintain power in elections (Klomp, 2020). In Türkiye, such dynamics might have conditioned voters’ interpretations through unequal access to state resources, reconstruction assistance expectations, or perceived political risks of opposing the incumbent. The results of the 2023 elections in favour of the government further support these mechanisms even though the elections occurred on an uneven playing field for the opposition and amid widespread speculation about potential vote manipulation. While our survey cannot model these mechanisms directly, their presence highlights the limits of standard retrospective assumptions in Türkiye’s political environment and helps explain why the February 2023 earthquakes’ political consequences may not align with classic electoral punishment and reward assumptions.
The media landscape and partisanship play especially significant roles in electorally autocratic contexts, which we can analyse through our data. Three factors demonstrate partisanship as the most substantial element in relation to incumbent evaluations. First, party support was associated with every aspect of government performance we could model, unlike personal loss or emotions. This highlights partisan bias above any other indicator. Second, the high odds ratios in the results for party support in all three models indicate that the magnitude of partisan bias was higher than any other factor. Finally, there is evidence to suggest that partisanship overlapped with the other indicators, with opposition supporters trusting social media and feeling negative emotions more (see the descriptive Table 2).
Voters across emotional dispositions and trust in social media.
This last point warrants emphasis from a trauma literature perspective. Beyond individual-level factors, social and political factors influence whether people experience disasters as traumatic. Being part of a community impacts emotions and perceptions. Theoretical discussions in the cultural trauma literature specifically draw attention to the influence of ‘carrier groups’ that frame and narrate events in ways that cause affective reactions (Alexander, 2004). Our findings provide empirical support for this theoretical framework. Table 2 demonstrates that opposition party supporters, who can be understood as a distinct political community following different news sources and leaders, had different general emotions than government supporters. They reported negative emotions more frequently and showed greater trust in social media than incumbent coalition voters. While we cannot determine reasons for negative emotions with available questions, the pattern aligns with collective trauma theory’s emphasis on how group membership shapes disaster interpretation.
These findings have further implications on how partisanship operates in electoral autocracies. The asymmetry in responsiveness across voter groups supports other scholarly work that demonstrates that partisan bias is stronger among government supporters (Aytaç et al., 2025). Such voters’ positive evaluations appear insensitive to evident hardships, corruption, and administrative mismanagement, which were visible in the earthquakes’ wake. Opposition voters, who reported higher levels of negative emotions (especially anxiety), likely sought out more information and were more willing to update their beliefs accordingly, whereas government supporters, experiencing fewer negative emotions, may have remained positively biased and resistant to revising their evaluations (Atkeson and Maestas, 2012; Weeks, 2015). Thus, partisanship operates less as a reciprocal bias affecting both camps equally than as a stabilising force sustaining incumbent support, even when material conditions would otherwise prompt reassessment. Overall, the most conclusive finding from this research is that individuals entrenched in persistent polarisation and a restricted media landscape had government evaluations in the aftermath of the disaster markedly associated with their pre-existing political views.
Conclusion
This article examined the relationship between government evaluations and political and social factors, including emotions, following natural disasters. Based on a public opinion survey conducted in Türkiye after the February 2023 earthquakes, we analysed the significance of disaster-related losses, emotions, political party support, and media trust for perceptions of the government’s disaster management, earthquake preparedness, and responsibility for losses. The results show that earthquake-related losses increased negative perceptions of the incumbent’s actions, in line with the attentive retrospective hypothesis. Negative emotions increased the likelihood of evaluating the government’s performance adversely, supporting the blind retrospective thesis. However, inconsistent patterns in blame attribution aligned more with attentive decision-making and the citizens’ ability to distinguish between different dimensions of government accountability.
Most significantly, affective polarisation emerged as the strongest and most consistent predictor across all dimensions of government evaluation. Opposition supporters were systematically more critical of government performance and substantially more likely to attribute blame. The findings demonstrate that in highly polarised contexts, pre-existing political loyalties may override both disaster experiences and emotional moods in shaping accountability judgements. This suggests that polarisation creates different interpretive frameworks determining how citizens process disasters and assign political responsibility.
These findings contribute to two key literatures. First, they advance scholarship on natural disasters and electoral outcomes by providing a clarification why, unlike the 1999 Marmara earthquakes that triggered electoral realignment, the 2023 earthquakes did not produce political change despite comparable devastation and government failures. Our analysis suggests that government supporters’ pre-existing partisan attachments biased how they interpreted the incumbent’s earthquake performance. Second, they contribute to research on disasters and regime change by demonstrating how affective polarisation in electoral autocracies can insulate incumbents from accountability that might otherwise prompt democratisation.
The findings stress the need to expand disaster politics literature beyond democratic contexts and differentiate between dimensions of government responsibility. As climate change intensifies disaster frequency, the Global South faces more catastrophic consequences due to economic and infrastructural limits, often compounded by media restrictions and constrained political freedoms. Three years after the Turkish earthquakes, approximately half of the 632,667 destroyed homes have yet to be reconstructed, and 360,455 individuals continue to reside in container housing. Socioeconomic indicators in the affected regions also remain markedly weaker, with unemployment rates higher and per capita income lower than the national average (Kahramanmaraş ve Hatay Depremleri Yeniden İmar ve Geliştirme Raporu, 2026). Although these statistics cannot fully capture lived experiences, they highlight that humanitarian crises following natural disasters are not brief emergencies but long-term processes whose consequences influence societies for years to come. Understanding how political factors such as polarisation relate with disaster accountability becomes even more critical when communities face such prolonged recovery periods, given that institutional and regime characteristics are likely to continue influencing reconstruction, resource allocation, and future disaster policies.
Supplemental Material
sj-docx-1-bpi-10.1177_13691481261447632 – Supplemental material for Perceptions of government performance and collective trauma: The case of the 2023 Turkish earthquakes
Supplemental material, sj-docx-1-bpi-10.1177_13691481261447632 for Perceptions of government performance and collective trauma: The case of the 2023 Turkish earthquakes by Yaprak Gürsoy and Buğra Güngör in The British Journal of Politics and International Relations
Supplemental Material
sj-docx-2-bpi-10.1177_13691481261447632 – Supplemental material for Perceptions of government performance and collective trauma: The case of the 2023 Turkish earthquakes
Supplemental material, sj-docx-2-bpi-10.1177_13691481261447632 for Perceptions of government performance and collective trauma: The case of the 2023 Turkish earthquakes by Yaprak Gürsoy and Buğra Güngör in The British Journal of Politics and International Relations
Footnotes
Acknowledgements
The authors would like to thank the anonymous reviewers for reading earlier versions of this article and providing constructive feedback. They also thank Luis Remiro, the participants of the European Consortium for Political Research (ECPR) Standing Group on Affective Polarization at the 2025 General Conference in Thessaloniki, and the Political Studies Association (PSA) Specialist Group on Turkish Politics at the 2025 Annual International Conference in Birmingham for their valuable comments.
Data Availability Statement
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Yaprak Gürsoy received the London School of Economics and Political Science (LSE) Urgency Fund on 20 April 2023 to purchase data from KONDA Research and Consultancy Inc. after the February 2023 Turkish earthquakes.
Supplemental material
Additional supplementary information may be found with the online version of this article.
SUPPORTING INFORMATION 1: Survey Questionnaire
SUPPORTING INFORMATION 2: Diagnostics and Robustness Checks
Table A4: Summary statistics
Table A5: Regression models using the question ‘In the June 24, 2018, general parliamentary elections, who did you vote for, which party did you give your vote to?’ for opposition party support variable
Table A6: Regressions with each negative emotion as a separate independent variable
Table A7: Coefficient (odds ratios) plots and margins plots
Table A8: Hierarchical regression models
Table A9: Missing observations in the models
Table A10: Distance calculated from the city of Hatay that suffered the most damage due to the earthquakes
Table A11: Check for outcome imbalance and separation bias for model 3: blame attribution and supporting opposition parties
Table A12: Partisanship and gender across the models
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
