The discourse surrounding democratic legitimacy involves a diverse array of researchers, with Easton’s pioneering studies (1975, 1965) on political support serving as a cornerstone reference (BORBA and CARDOSO, 2021)⁴. The transition toward analyzing legitimacy from a multidimensional perspective began in the early 1990s, and despite past empirical limitations, the contemporary proliferation of public opinion data has facilitated the measurement of legitimacy⁵. Researchers have increasingly adopted multidimensional approaches, with Norris (2011, 1999) and Dalton (2004) making notable contributions to the reformulation of Easton's framework (BORBA and CARDOSO, 2021). For the analyses conducted in this article, we focus on Booth and Seligson's (2009) multidimensional conception, which also aligns with Easton’s framework. While Booth and Seligson (2009) share Norris' (2011, 1999) multidimensional perspective on political support, they have expanded upon it by utilizing the concept of ‘democratic legitimacy’ rather than ‘regime support’⁶.

The book ‘The Legitimacy Puzzle in Latin America’ by Booth and Seligson, published in 2009, is highly relevant to both empirical and theoretical discussions on political legitimacy in Latin America (RENNO et al., 2011). In their study, the authors contend that fundamental questions about legitimacy remain unanswered, focusing on what they perceive as the major puzzles: the structure, sources, and effects of legitimacy. Thus, in their analyses of Latin America, the authors identified six dimensions of the structure of democratic legitimacy⁷, ranging from the most diffuse to the most specific.

Booth and Seligson's empirical approach (2009) to examining support for democracy focuses more on citizens’ endorsement of democratic norms and practices rather than their preference for abstract types of regimes. In this context, the authors highlight the problem of social desirability, which may hinder respondents from expressing support for dictatorships. Furthermore, surveys have revealed that different audiences have different interpretations regarding items related to democracy. “As noted by Magalhães (2018), this approach, which avoids directly addressing the type of regime in survey questions, has been widely adopted” (BORBA and CARDOSO, 2021, p, p. 340). Additionally, the ‘focus on regime principles’ has led to the recognition that preferences for democracy may not be deeply ingrained, often coexisting with less liberal views⁸.

The multidimensional approach has made significant strides by challenging more explicit and direct approaches to democracy, understanding legitimacy in a broader sense while also considering empirical operationalization across a range of variables. It is important to remember here that explicit approaches seek to measure democracy in comparison to non-democratic regimes, encompassing questions such as “democracy is preferable to any other form of government” and “while democracy may have its flaws, it remains the best form of government”. As a result, these approaches directly refer to democracy, assuming that respondents define it in the same way (CARLIN, 2018, p. 401).

The multidimensional approach to democratic legitimacy, as developed by Booth and Seligson (2009), has had significant implications for Latin American researchers, particularly among Brazilians. It's important to note that these analyses do not uniformly adopt Booth and Seligson’s propositions (2009),

Considering these factors, we note that we will employ a multidimensional approach in our analysis of Latin American countries. Nonetheless, we also acknowledge the criticisms leveled against multidimensionality, which include: challenges in accurately diagnosing democracy; lack of consensus among researchers regarding which indicators to use; gaps in formulating more direct measures of democratic beliefs; and limitations in comparative analyses between countries and in developing longitudinal analyses (CARLIN, 2018; FERRÍN and KRIESI, 2016; JONGE, 2016; RENNÓ et al., 2011). Hence, we will reassess our appropriation of the multidimensional approach accordingly, concentrating on the three dimensions that Booth and Seligson (2009) identify as pivotal to the erosion of democracy: regime principles, support for regime institutions, and evaluation of regime performance. The authors argue that dissatisfied citizens in these three dimensions may display political or electoral behaviors that affect democratic stability.

Therefore, our revised analysis of democratic legitimacy developed in this article focuses on what the authors themselves deem central for democratic stability: 01. regime principles, and 02. support for regime institutions. Regarding regime performance, we believe that exploring it as an independent variable may be more effective than necessarily integrating it as a dimension of legitimacy itself. For the countries of Latin America, we utilized data from the LAPOP for 2016/2017 and 2018/2019, employing multilevel analysis to identify whether (and which) contextual factors of the countries play a role in the relationship between fear of crime, victimization, and democratic legitimacy.

Although democratic legitimacy and satisfaction with democracy are analytically distinct dimensions (TORCAL and MONCAGATTA, 2011), both are known to be related to the evaluation of government performance and/or presidential popularity (DALTON, 2004). Recent studies (MAZEPUS and TOSHKOV, 2022; SVOLIK, 2020) suggest that voters who are highly satisfied with the economy or sympathetic to the ruling party are more willing to accept violations of checks and balances, indicating that democratic support is more contingent than initially predicted in the theory (EASTON, 1975).

In Latin America, researchers have dedicated considerable attention to the issue of victimization, especially in studies that explore the connection between violence and political behavior. This issue is closely tied to the firsthand experience of violence in a region that experiences high homicide rates. Conversely, research conducted in other contexts, such as European countries, has underscored the significance of fear of crime and its correlation with widespread social insecurity (FARRALL, JACKSON and GRAY, 2009; VALENTE and PERTEGAS, 2018).

The literature underscores the importance of recognizing fear of crime and victimization as distinct yet interconnected phenomena, each potentially influenced by different causal mechanisms (ALTAMIRANO, BERENS, and LEY, 2020; MALONE, 2012). Notably, individuals feeling insecure tend to resort more frequently to private means of protection in response to heightened perceptions of insecurity. Furthermore, individuals affected by crime may be more inclined to urge government intervention and advocate for the implementation of public policies to tackle the diverse costs associated with crime, as well as the emerging needs and challenges stemming from violence (ALTAMIRANO, BERENS, and LEY, 2020).

Regarding the effects of violence on democratic legitimacy, Fernandez and Kuenzi (2010)¹⁰ observed that perceptions of crime and insecurity influence attitudes toward democracy in the Latin American context. They suggest that individuals who feel safer are generally more satisfied with democracy as a form of government. Carreras (2013), on the other hand, focused on diffuse support as a measure of democratic legitimacy, influenced by Easton’s conceptualization of political support (1975). This perspective aligns closely with the multidimensional concept of legitimacy developed by Norris (1999) and Booth and Seligson (2009), as discussed earlier. Therefore, Carreras (2013) argues that victimization and perceptions of insecurity have a negative impact on political support, posing a significant threat to the quality of democracy in the region.

In general, studies suggest that in Latin America, victimization and fear of crime have a detrimental impact on trust in political institutions from a comprehensive perspective. Consequently, this leads to a widespread decline in trust in democratic institutions, particularly evident in institutions directly involved in crime administration, such as the police and the criminal justice system (DAMMERT, 2012; BREÑA, 2019; SILVA and RIBEIRO, 2016). Consequently, the inefficiency of these institutions may help explain the decline in democratic legitimacy in the region, serving as a potential causal mechanism¹¹.

Building upon the discussion outlined in this section, we will proceed to present the analyses for Latin American countries utilizing data from two waves of the Americas Barometer – LAPOP from 2016/2017 and 2018/2019. Our analyses will center on the multidimensional conception of democratic legitimacy and explore the influence of contextual factors.

Between 2016 and 2018, the majority of Latin American countries experienced a decline in democracy dimensions as assessed by the Varieties of Democracy (V-Dem), with the liberal and participatory dimensions being particularly affected¹². Our analysis of the influence of homicide rates on democracy indicators (Polity IV and V-Dem) revealed how violence contributed to this decline¹³, further reinforcing our aim in this article to examine the relationship between violence and democratic legitimacy.

To analyze the impact of victimization and fear of crime on the decline of democratic legitimacy, we relied on data from the Americas Barometer, conducted by the Latin American Public Opinion Project (LAPOP)¹⁴ affiliated with the Vanderbilt University. Since 2004, LAPOP has been conducting regular opinion surveys on democratic values, behaviors, and socio-economic conditions across countries in Latin America, Canada, and the United States. In 2016, the survey covered 29 countries, totaling 43,000 interviews, while in 2018, the study was conducted in 20 countries with 31,050 respondents. Notably, the survey includes questions about victimization and fear of crime, as well as questions aligned with the multidimensional approach to democratic legitimacy.

Victimization has been a recurring theme in the survey, particularly through the question: ‘Have you been a victim of any type of crime in the past twelve months? That is, have you been a victim of robbery, burglary, assault, fraud, blackmail, extortion, violent threats, or ‘any other type of crime’ in the past 12 months?’ This question specifically pertains to direct victimization, meaning that the actual respondent has been the victim of a crime, and does not encompass indirect victimization (when another family member has been a victim).

The question on fear of crime is framed as follows: ‘Speaking of the place or neighborhood where you live, and thinking of the possibility of falling victim to an assault or a robbery, do you feel very safe, somewhat safe, somewhat unsafe or very unsafe?’. In turn, questions regarding changes in behavior resulting from the fear of victimization were exclusively carried out in El Salvador and Honduras during the 2018 survey wave.

We must also mention the significant lack of research on victimization in Latin America¹⁵, particularly in terms of comparative studies across countries addressing topics such as democratic legitimacy and other political dimensions. In the Latin American context, aggregated data on this topic mainly come from opinion surveys such as Latinobarómetro and the Americas Barometer (LAPOP). However, these two institutions employ different methodologies and approaches in their questions about fear of crime and victimization, posing challenges for cross-comparative analysis (DAMMERT and TOBAR, 2018). Despite LAPOP's limitations in formulating questions about fear of crime, it remains the most comprehensive and suitable survey for addressing our research objectives in Latin American countries.

Studies that specifically address fear of crime in South American countries have mainly focused on the consequences of feeling of insecurity in both public and private life, rather than delving into the underlying causes of this phenomenon. This may partly stem from an assumption that correlates fear of crime with high crime rates (VALENTE and VACCHIANO, 2020). Moreover, there is a scarcity of comparative studies on the causes of fear of crime in this region, as studies tend to concentrate on individual countries or specific regions, based on isolated victimization surveys.

In any case, existing studies suggest that fear of crime takes on unique characteristics in Latin American countries¹⁶. When comparing Argentina and Brazil, Valente and Macchiano (2020) argued that in Argentina, fear of crime is also influenced by non-criminal aspects, in addition to being a phenomenon linked to other forms of social insecurity. In contrast, in Brazil, victimization emerges as the predominant explanation, following a more traditional pattern where fear of crime is mainly rooted in direct experiences of violence.

Given our research problem, which focuses on the impacts of different dimensions within the national context (such as political violence, crime, and the political-institutional dimension of democratic quality), we have opted for a research design that integrates individual-level data (from surveys) with country-level data (from various sources described below), made possible through the use of multilevel analysis techniques¹⁷.

The database consists of two levels. Level 01 encompasses data from the 2016 and 2018 LAPOP surveys for seventeen Latin American countries: Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, El Salvador, Ecuador, Guatemala, Honduras, Mexico, Nicaragua, Panama, Paraguay, Peru, the Dominican Republic, and Uruguay. We excluded Haiti and Venezuela from this stage of the research as the LAPOP survey was not conducted in these countries in 2018.

Our dependent variables (DVs) consist of two measures of democratic legitimacy derived through data reduction techniques, more specifically factor analysis¹⁸, utilizing questions traditionally employed in studies conducted within the framework of a multidimensional approach, as described above (BOOTH and SELIGSON, 2009; MAGALHÃES, 2018). A detailed description of the questions is provided in the appendix. The ‘first factor’ aggregates the following variables: trust in congress, trust in political parties, trust in the President, and trust in elections. The ‘second factor’, in turn, aggregates the following variables: respect for institutions, pride in the political system, support for the political system, and protection of basic rights.

Subsequently, we aggregated the variables from Factor 01 to create the Institutional Trust Index (ITindex)¹⁹, recodified on a scale of 01 to 10. In this scale, 01 represents the lowest trust in institutions, while 10 denotes the highest trust. The Cronbach’s alpha coefficient for the index was calculated at 0.797. Notably, among the variables comprising the institutional trust index, trust in elections and the president exhibited the highest averages (5.424 and 5.039, respectively), while trust in political parties displayed the lowest average (3.834) among Latin American countries.

We also aggregated the variables from Factor 02 to create the Diffuse Support Index (DS Index)²⁰. This index was similarly recoded on a scale from 01 to 10, where 01 indicates the lowest level of diffuse support, and 10 represents the highest. The Cronbach’s alpha for this index was calculated to be 0.791. Notably, among the variables comprising the diffuse support index, respect for institutions and support for the political system showed the highest values across countries (averages of 6.511 and 6.276, respectively), while protection of basic rights exhibited the lowest average (3.481).

Table 01 displays the median and average values of the institutional trust index and the diffuse support index. Furthermore, for comparative purposes, we incorporated the ‘Churchillian’ measure of adherence to democracy in the same table. The diffuse support index has an average of 5.829, indicating higher values compared to the institutional trust index, which stands at 4.8. This suggests that, overall, support for democracy receives a better evaluation among Latin Americans.

Table 01

	Institutional Trust Index		Diffuse Support Index		Adherence to Democracy (Churchillian)
Country	Median	Average	Median	Average	Median	Average
Argentina	4.643	4.746	6.071	5.821	8.571	7.861
Bolivia	5.357	5.148	6.071	5.923	7.143	6.479
Brazil	3.929	4.230	4.643	4.640	7.143	6.673
Chile	4.643	4.624	5.714	5.403	7.143	7.086
Colombia	4.286	4.492	6.071	5.848	7.143	6.745
Costa Rica	5.357	5.151	7.143	6.902	8.571	7.607
El Salvador	4.643	4.725	6.071	5.927	7.143	6.655
Ecuador	5.357	5.246	6.429	6.236	7.143	6.512
Guatemala	4.643	4.626	6.071	6.107	5.714	6.315
Honduras	4.286	4.378	5.714	5.421	5.714	6.170
Mexico	5.000	4.924	6.071	5.944	7.143	6.605
Nicaragua	5.714	5.502	6.786	6.536	7.143	6.569
Panama	3.929	4.176	5.714	5.654	7.143	6.687
Paraguay	4.643	4.672	5.357	5.408	5.714	6.365
Peru	4.286	4.409	5.714	5.454	7.143	6.399
Dominican Republic	5.000	4.954	5.714	5.601	7.143	6.869
Uruguay	5.714	5.801	6.786	6.413	8.571	8.207
Latin America	4.643	4.800	6.071	5.829	7.143	6.802

Source: Elaborated by the authors based on LAPOP 2018-2016.

The Brazilian case stands out, displaying the lowest values for both the institutional trust index (4.230) and the diffuse support index (4.640) – the latter being significantly lower than the regional average (5.829). However, adherence to democracy in ‘Churchillian’ terms, i.e., the belief that democracy is the best form of government, shows a considerably higher value (6.673), close to the Latin American average.

At level 01, we considered the following independent variables: victimization, feeling of insecurity, interpersonal trust, evaluation of the country’s economic situation, and individual assessment of one's own economic situation. The control variables included gender, age groups, education level, race, and region of residence. For detailed descriptions of the variables, please refer to Table 03 in the Annexes.

To construct the level 02 data (context), we incorporated the following variables: 01. the country's homicide rate (moving average)²¹; 02. the level of political stability and absence of violence²²; and 03. democracy indicators (Polity IV and V-Dem). Each country was duplicated in this dataset to provide respective information for each period. Since the level 01 data stem from two survey waves (LAPOP 2018, 20, 2016), each country/year combination became a case at level 02. In other words, the three mentioned variables above were collected for both 2016 and 2018, resulting in an N of 34 cases at level 02.

The most appropriate statistical model for analyzing this data was multilevel modeling, which enables us to determine the direct effects of individual and contextual explanatory variables. Additionally, it allows us to evaluate whether explanatory variables at the macro level moderate relationships at the individual level, indicating potential variability in the impact of a predictor based on others (HOX, 2010).

The technique is particularly valuable for examining the relationship between individuals and their collectives, operating under the assumption that people interact within and are influenced by the groups or social contexts to which they belong. Therefore, it is structured as a hierarchical system (RAUDENBUSH and BRYK, 2002; SOMMET and MORSELLI, 2017).

To achieve this, the models are structured with the response variable positioned at the lowest level, i.e., assuming the existence of a dataset with a single outcome or response variable measured at the first level, alongside explanatory variables at all levels. We employed fixed-effects models to refine the focus of our analyses. However, during the exploration and definition of the research design, we also conducted random-effects models for victimization and feeling of insecurity. This enabled us to more clearly identify variations in the effect of victimization among countries, a phenomenon not observed to the same extent for feeling of insecurity, indicating greater homogeneity regarding the latter variable across Latin America

Our research hypotheses considered the contextual factors of countries based on three main dimensions: the country’s homicide rate (H.01); the degree of political stability and absence of political violence (H.02); and democracy indicators in the country (Polity IV and V-Dem²³) (H.03)²⁴. They are as follows:

H.01 – The homicide rate in the country does not have a direct impact on democratic legitimacy, but rather operates through cross-level interactions between the micro and macro levels; — Contextual effect – violence²⁵;

H.02 – The impact of victimization on democratic legitimacy diminishes in contexts with greater political stability and absence of violence; — Contextual effect – political violence;

H.03 – The impacts of victimization and fear of crime on democratic legitimacy are more pronounced in less consolidated democracies; — Contextual effect – democracy indicators; this hypothesis suggests that less consolidated democracies are more susceptible to being affected by violence²⁶.

Due to space limitations, this section focuses solely on the tests conducted using the ‘institutional trust index’ as the dependent variable. However, we also mention our second dependent variable in the main text, with the detailed results of these tests available in the appendix.

Initially, we conducted a multiple regression analysis for the institutional trust index. The model²⁷ explains 9.2% of the variance in the institutional trust index. Being a victim of crime implies a reduction of 0.31 in the index, even when other variables are controlled. Feeling insecure implies a reduction of 0.28 in the index. Overall, the variable with the greatest influence on the model was the assessment of the country’s economic situation, which reduces the index by -0.89. Additionally, victimization has a similar effect to individuals’ assessment of their own economic situation. Older individuals demonstrate higher levels of trust in institutions, whereas those with higher education levels tend to exhibit more distrust. Moreover, there is a noticeable similarity between the impact of victimization and education level in the model.

Table 02 presents the models developed for the multilevel analysis²⁸, delineating each of the hypotheses proposed. As noted previously for the multiple regression analysis, victimization and feeling of insecurity exert a negative impact on institutional trust. However, the multilevel analysis enables the inclusion of the aforementioned contextual variables. Our selection of variables considered the indicators used by Carreras (2013) and Fernandez and Kuenzi (2010), but we also explored dimensions not covered in these studies, such as homicide rate, political stability, and absence of political violence.

Table 02

Variables	Hypothesis Model 01 - with interaction between variables	Hypothesis Model 02	Hypothesis Model 02 – with interaction between variables	Hypothesis Model 03 – with interaction between variables	Hypothesis Model 03 – with interaction between variables
Victimization	-0.265*** (0.030)	-0.331*** (0.021)	-0.436*** (0.056)	-0.696*** (0.080)	-0.331*** (0.021)
Feeling of Insecurity	-0.285*** (0.019	-0.285*** (0.019	-0.286*** (0.019)	-0.285*** (0.019)	-0.600*** (0.085)
Gender	0.083*** (0.018)	0.084*** (0.018)	0.084*** (0.018)	0.083*** (0.018)	0.0836*** (0.0184)
Age (age bracket)	0.107*** (0.007)	0.107*** (0.007)	0.107*** (0.007)	0.107*** (0.007)	0.107*** (0.007)
Educational Level	-0.421*** (0.026)	-0.421*** (0.026)	-0.421*** (0.026)	-0.418*** (0.026)	-0.420*** (0.026)
Interpersonal Trust	-0.255*** (0.020)	-0.254*** (0.019)	-0.255*** (0.019)	-0.254*** (0.019)	-0.255*** (0.019)
Country’s Economic Situation	-0.820*** (0.021)	-0.821*** (0.021)	-0.821*** (0.021)	-0.820*** (0.021)	-0.821*** (0.021)
Individual’s Economic Situation	-0.340*** (0.021)	-0.340*** (0.021)	-0.340*** (0.021)	-0.340 (0.021)	-0.340*** (0.021)
Homicide Rate (moving average)	-0.006 (0.005)
Homicide rate (moving average)* Victimization	-0.004** (0.001)
Political stability and Absence of violence		0.010* (0.005)	0.009 (0.009)
Political stability and absence of violence* Victimization			0.002* (0.001)
V-Dem Participation				-0.135 (0.665)
V-Dem Participation*Victimization				0.784*** (0.166)
V-Dem Polyarchy					-0.608 (0.556)
V-Dem Polyarchy* Feeling of insecurity					0.462*** (0.121)
AIC	198889.4	198883.2	198892.7	198858.8	198867.4
BIC	199003.3	198988.4	199006.6	198972.7	198981.3
Observations	47224	47224	47224	47224	47224
Groups	34	34	34	34	34

Source: Elaborated by the authors based on LAPOP 2018-2016.

Note: p<0.1 * p<0.05 ** p<0.01 *** p<0.001.

Therefore, we conclude that Research Hypothesis 01 has been confirmed, as victimization does not directly interfere with democratic legitimacy but rather operates through interaction with the moving average homicide rate. Additionally, Hypothesis 02, concerning the impact of the contextual factors of political violence and political stability, has been confirmed, both independently and in interaction with victimization. Regarding Hypothesis 03, which examines the effects of victimization and fear of crime across different democratic contexts, it is partially confirmed, as democracy indicators showed significance only through interaction. Specifically, Polity IV interacted solely with victimization, while all V-Dem indicators tested interacted with both victimization and feeling of insecurity.

To offer a more detailed examination of the interactions depicted in the table, we computed the predicted values of the institutional trust index for both crime victims and non-victims. In Graph 01, the dashed line illustrates the impact of victimization in a context with lower homicide rates, while the solid line depicts the effect of victimization in a context with higher homicide rates.

Graph 01 Source: Elaborated by the authors based on LAPOP 2018-2016.

Therefore, we find that in contexts with high homicide rates, trust in institutions tends to be slightly lower. Consequently, elevated homicide rates exacerbate the impact of victimization, leading to a decline in the institutional trust index. These findings underscore our assertion concerning the influence of violence on democratic legitimacy. Not only does victimization adversely affect trust in institutions, but nations with higher homicide rates also magnify the impact of being a crime victim.

To offer a more detailed examination of the interactions depicted in the table, we computed the predicted values of the institutional trust index for both crime victims and non-victims. In Graph 01, the dashed line illustrates the impact of victimization in a context with lower homicide rates, while the solid line depicts the effect of victimization in a context with higher homicide rates.

Therefore, we find that in contexts with high homicide rates, trust in institutions tends to be slightly lower. Consequently, elevated homicide rates exacerbate the impact of victimization, leading to a decline in the institutional trust index. These findings underscore our assertion concerning the influence of violence on democratic legitimacy. Not only does victimization adversely affect trust in institutions, but nations with higher homicide rates also magnify the impact of being a crime victim.

Hypothesis 02 investigated the impact of violence in conjunction with political processes by incorporating the World Bank indicator assessing perceptions of the likelihood of political instability and politically-motivated violence. It is noteworthy that this indicator holds significance both independently and in interaction with victimization, underscoring the relevance of the connection between violence and political factors.

In Graph 02, the solid line represents a higher level of political stability and absence of violence. In more politically stable countries, the institutional trust index tends to be higher (5.28) compared to less stable countries (5.05). Furthermore, the impact of victimization on trust in institutions is more pronounced in contexts of political instability and with higher levels of political violence. More specifically, the analysis of predicted values indicates a difference of 0.60 between an individual who has not been a victim of a crime in a country with high stability and an individual victim of a crime in a country with low stability.

Graph 02 Source: Elaborated by the authors based on LAPOP 2018-2016.

We did not confirm this same hypothesis for the diffuse support index, possibly because the institutional trust index is more susceptible to short and medium-term conjunctural changes. Nevertheless, we should consider this finding in light of the arguments presented by Ley (2018), who emphasizes the role of violence in reshaping the broader political landscape, which could lead to a decline in voter turnout due to the heightened costs and risks associated with voting. Additionally, organized crime often develops political interests, above all to secure economic benefits and operate with relative impunity, and elections serve as a crucial mechanism for restructuring these protective networks (LEY, 2018).

Hypothesis 03 posited that victimization and fear of crime would exert more pronounced effects on democratic legitimacy, in less consolidated democracies, suggesting that these democracies, would be more vulnerable to the impact of violence. This hypothesis was partially confirmed insofar as democracy indicators became significant when interacting with victimization and feeling of insecurity. Regarding the interaction between victimization and V-Dem Participation, we found that more democratic contexts diminish the impact of being a victim of a crime on institutional trust.

We also conducted multilevel analysis for the diffuse support index, confirming the negative impacts of victimization and feeling of insecurity on democratic legitimacy. Hypothesis 01, pertaining to the cross-level interaction between victimization and the violent contextual environment (moving average homicide rate), was confirmed. This reinforces the association between the concrete experience of victimization at the micro level and the country’s homicide rate at the macro level. Specifically, an individual in a context with a low homicide rate who has not been a victim of a crime presents a diffuse support index of 6.17, while an individual in a context with a high homicide rate who has been a victim of a crime presents a diffuse support index of 5.69. In other words, a high homicide rate amplifies the effect of victimization, leading to a decrease in the diffuse support index.

Hypothesis 02, which addressed political stability and the absence of violence, was not confirmed for the diffuse support index, unlike what we found for the institutional trust index. Hypothesis 03, concerning the pronounced effects of victimization and fear of crime on democratic legitimacy in less consolidated democracies, was partially confirmed. While the democracy indicators individually did not show significance in the model, significance emerged through the cross-level interaction with victimization for the V-Dem indicators. In a more democratic context (in terms of participation), we find a diminished impact of victimization on diffuse support. This finding reinforces the notion that less consolidated democracies tend to experience a greater impact from violence (LAFREE and TSELONI, 2006).

Overall, our analyses underscored the detrimental effects of victimization and fear of crime on the decline of democratic legitimacy in Latin America, consistent with findings from the literature. Specifically, our study reinforces Fernandez and Kuenzi’s (2010) findings regarding the significance of victimization and their critique of democratic performance. However, this study advances beyond Carreras (2013) and Fernandez and Kuenzi (2010) by considering the impact of the violent context, based on the moving average homicide rates for the countries under study. Thus, we may assert that high homicide rates exacerbate the negative impact of victimization on the decline of democratic legitimacy.

While victimization and fear of crime certainly impact the decline of democratic legitimacy, our multilevel analysis revealed partially differentiated interactions with contextual variables. This finding reinforces our argument that these phenomena have distinct consequences, suggesting a connection with democratic legitimacy that does not necessarily operate through the same causal mechanisms.

Upon examining democratic legitimacy among Latin Americans, we found that support for democracy in broader terms, termed as diffuse support, was higher compared to trust in institutions. However, Brazil stood out, recording the lowest values for both the institutional trust index and the diffuse support index, despite also showing high levels of support for democracy in the ‘Churchillian’ sense. Insights from Cohen et al. (2021) shed light on these findings, suggesting that Bolsonaro’s victory may have generated short-term satisfaction, particularly among critics of democracy, potentially fueling a ‘reservoir of support’ for future democratic violations. This case underscores the importance of employing multidimensional measures of legitimacy beyond traditional democracy indicators.

When investigating the influence of violence on democratic legitimacy, we observed that being a victim of crime and feeling insecure reduce both the institutional trust index and the diffuse support index, with the latter being slightly more impacted. Victimization, in particular, exerts an influence on diffuse support similar to other variables commonly studied in political behavior literature, such as interpersonal trust and economic assessment (in the egotropic dimension), thereby reinforcing Carreras' (2013) findings. Overall, the analyses conducted for our article confirm the detrimental effects of victimization and fear of crime on democratic legitimacy in Latin America.

We selected indicators for the multilevel analysis based on criteria outlined by Carreras (2013) and Fernandez and Kuenzi (2010), while also exploring aspects not addressed in these studies, such as the homicide rate, political stability and absence of political violence. Our multilevel analysis findings suggest that in contexts with high homicide rates, both trust in institutions and diffuse support tend to be lower.

In other words, while homicide rates alone do not show significance in legitimacy indices, this correlation emerges when interacting with victimization, indicating that this variable is better suited to focus on specific violent contexts. This dimension, overlooked in previous studies, could refine the use of violence indicators for analyzing political phenomena. These findings confirm our argument regarding the impact of violence on democratic legitimacy, as countries with high homicide rates experience an intensification of the effect of being a victim of a crime. Furthermore, analyses using V-Dem indicators reinforce the argument that less consolidated democracies tend to be more affected by violence (LAFREE and TSELONI, 2006).

The impact of victimization on institutional trust becomes particularly pronounced in contexts marked by political instability and heightened levels of political violence. This finding echoes Ley’s (2018) emphasis on the role of violence – such as homicide rates, organized crime activity, and violence targeting political candidates and parties – in reshaping the broader political landscape. As such, political instability and violence also carry far-reaching consequences for democratic legitimacy.

Our article does not aim to conclusively settle the debate on victimization, fear of crime, and democratic legitimacy. Instead, we aim to contribute to advancing future studies by pointing to different methodological and theoretical frameworks. However, we stress the importance of giving primary consideration to the issue of violence, in its various forms, in Latin American studies on political behavior, rather than treating it as a marginal variable. Labeling a country as democratic while it experiences homicide rates comparable to war zones, or overlooking the impact of punitive discourses and abuses committed under the guise of security and fear, means embracing a conception of democracy that inherently contradicts itself.