Perceptions of Security, Victimization, and Coexistence: A Database from Cali, Colombia

Abstract

This article addresses a key evidence gap in urban safety policy in Colombia: the absence of publicly accessible microdata that jointly measure victimization, perception of security, and probability of sanctions among socioeconomically vulnerable residents. It aims to provide a clean, linkable dataset that enables analysis of variations in these issues across demographic and territorial groups in Cali (recently classified as the 29th most dangerous city worldwide, with 1028 and 1065 homicides in 2024 and 2025, respectively). It reports face-to-face survey data collected from 22 July to 16 August 2024, at Sistema de Identificación de Potenciales Beneficiarios de Programas Sociales (SISBEN) service points. The final dataset includes 2139 adults (aged 18–95 years) and combines (i) primary responses on perceived safety (e.g., public space safety and surveillance cameras), perceived likelihood of sanction, victimization, and self-protection measures with (ii) selected sociodemographic and household characteristics drawn from SISBEN IV records. Individual-level linkage was implemented using respondent identification at interviews, yielding an integrated anonymized file suitable for replication and secondary analysis. The dataset enables distributive analyses of insecurity (e.g., by sex, age, and ethnicity—including Afro-descendant populations) within a policy-relevant target group and supports evaluation and targeting of local interventions by providing individual-level indicators.

Dataset: https://doi.org/10.17605/OSF.IO/FPB7G

Dataset License: Academic Free License (AFL) 3.0

1. Summary

Security is a fundamental component of quality of life and well-being, whereas violence reverses the development of sectors such as education, health, employment, and infrastructure provision [1]. Moreover, the physical security of property and individuals, along with perceptions of security, are vital to the analysis on poverty [2]. Alternatively, the theory of fear of crime has been developed on the basis of the distinction among objective victimization risk, perceived risk, and fear-related emotions, demonstrating that fear constitutes a specific phenomenon that cannot be explained solely by crime rates but also by social and symbolic processes associated with perceived vulnerability, the physical environment, and institutional trust [3,4,5,6,7,8]. In this regard, the literature noted that studies frequently associate fear with perceived risk, although they are distinct constructs with partially different determinants [4,5,6]. Similarly, contemporary approaches posit that fear of crime also operates as an expressive indicator of social order, in that it can reflect perceptions of institutional control, community cohesion, and environmental cues of disorder beyond direct victimization experiences [7,9,10].

This literature documents that victimization, fear of crime, and perceptions of insecurity are systematically associated with low levels of subjective well-being, life satisfaction, and social trust, which reinforces the concept that fear of crime is a central component of quality of life in urban areas [11,12,13]. In Latin America, analyses emphasize that fear can persist despite fluctuations in certain crime indicators due to its links with inequality, institutional weakening, conditions of the urban environment, and everyday experiences of violence [12,13,14]. In Colombia, evidence points to consistent patterns: victimization and perceptions of insecurity are associated with low levels of life satisfaction and high levels of social distress, which implies that subjective security directly influences well-being beyond recorded crime [15].

Cali represents a customary presence of insecurity. For example, the city was ranked 29th and 31st in 2024 and 2023, respectively, among the top 50 “most dangerous cities in the world” [16]. Indeed, the localization of homicides has been persistently concentrated within similar neighborhoods, such as within three comunas 1—14, 15, and 21—which have large proportions of populations registered in Sistema de Identificación de Potenciales Beneficiarios de Programas Sociales [SISBEN] 2) and comprise 31% of the cases recorded in 2024 [17]. In addition, studies that conducted causal evaluations of security policies focused on objective indicators of violence, such as juvenile curfews, and demonstrated that these interventions do not necessarily generate average reductions in homicide rates and may exhibit heterogeneous effects according to territorial and socioeconomic characteristics. This type of evidence underscores the limitations of an approach that is exclusively based on recorded crime and reinforces the importance of measuring the subjective and institutional dimensions of insecurity [18].

Therefore, the current study designed a survey that includes questions on perceptions of human security in urban areas as a part of the framework of Inter-Administrative Agreement No. 4132.010.32.1.376-2024, which was signed between the Fundación Universidad del Valle and the Departamento Administrativo de Planeación Distrital (DAPD) of the Alcaldía de Cali. This survey aims to address gaps in the metrics for perceptions on security within Cali, which are typically derived from aggregate records or private initiatives, such as Cali Cómo Vamos program [19], and do not offer public access to individual data, which is a disadvantage.

This article provides a particularly valuable empirical input, because it simultaneously measures perceptions of safety at the city and neighborhood levels; evaluations of strategies implemented by the Alcaldía de Cali—such as Espacios Públicos Seguros and Vigilancia y Prevención del Delito, as reported in [20]—as well as perceptions of the police, probability of sanction, household victimization experiences, and self-protection measures, together with a broad sociodemographic module and location identifiers at the comuna and neighborhood levels.

This structure enables the operationalization of components highlighted by fear of crime theory, such as perceived insecurity, perceived risk, perceived state effectiveness, institutional control, and self-protective behaviors, and enables the testing of hypotheses on their unequal distribution according to gender, socioeconomic status, education, or labor-market attachment [3,4,9]. Additionally, as [11,12,15] is linked to the SISBEN IV administrative database and focuses on socioeconomically vulnerable populations, the dataset supports a research agenda that examines the interconnections among fear of crime, poverty, and subjective well-being, consistent with evidence that documents the negative effects of victimization and fear on life satisfaction and social trust within the contexts of Latin America in general and Colombia in particular. In particular, the ability to estimate models that elucidate relationships among victimization, perceived safety, and material living conditions enables more refined tests of mechanisms proposed by the theory—such as the mediating role of fear in the relationship between victimization and well-being—and identifies high-vulnerability profiles that can inform policies that target citizen security in urban contexts such as Cali [3,4,7,12].

While featuring certain behavioral and vulnerability dimensions (e.g., specific avoidance practices and heterogeneity in trust across security providers—including the police, private security, and community networks) and individual attributes (e.g., disability or sexual orientation), the proposed instrument provides a solid empirical basis for the economic analysis of perceived insecurity and open the path for recipients of the more explicit incorporation of these element in future applications. The joint availability of measures of perceived safety at multiple territorial scales, household victimization experiences, indicators of self-protection, evaluations of institutional performance, and a detailed sociodemographic and spatial module could enable researchers to estimate associations and distributional gradients in perceived insecurity and explore relevant mechanisms such as relationships among victimization, institutional trust, and protective behaviors. In this manner, the study establishes an appropriate platform for future extensions without undermining the analytical value of the dataset.

This data repository enables researchers to explore the construction, perception, and narration of insecurity, victimization, and coexistence in urban environments. Questions on sense of belonging to a community and sense of security in various public and private spaces open avenues for studies on cultural values and social representation, all of which are key topics under the interdisciplinary humanities approach.

In addition, this dataset can be combined with other administrative records, such as those on health, education, and justice, which significantly expands its utility for the integrated analysis of insecurity conditions and their multidimensional impacts. Integration with administrative records fosters a comprehensive approach that links perceptions and experiences of victimization with structural outcomes, including the identification of spatiotemporal patterns that enrich the design of targeted, evidence-based public policies. Finally, the survey and database enable an inter-administrative agreement to accomplish its main objective—to promote the advancement of public policy and social initiatives that effectively utilize public funding, which, in this case, are grounded in the best use of statistical data.

2. Data Description

Table 1. Database details.

General areas	Sociology, economics, social policy, and human geography
Specific areas	Security, crime, victimization, and perception
Type of data
Method of data acquisition	Data were randomly collected through in-person surveys in Santiago de Cali, Colombia, from July to August (2024), at various SISBEN service points. The respondents, residing in 22 administrative divisions, contributed to a dataset that integrates primary survey data with SISBEN IV survey data, yielding a total of 2139 individual responses.
Data source location	Country: Colombia, Alcaldía de Cali
Data accessibility	Repository name: Open Science Framework Data identification number: DOI: 10.17605/OSF.IO/W6Y8X Direct URL: https://doi.org/10.17605/OSF.IO/FPB7G

presents the key topics included in the database at the comuna and neighborhood levels.

The database is divided into several sections: sociodemographic characteristics, perceptions of security and policy, sense of security at places, victimization (sanction), and self-security measures.

2.1. Socioeconomic Characteristics

Table 2. Summary of sociodemographic characteristics (n = 2139).

Variable	Category	N	(%)
Age group (years) *	18–20	117	5.47
	21–24	156	7.29
	25–29	258	12.06
	30–34	262	12.25
	35–39	274	12.81
	40–44	199	9.30
	45–49	173	8.09
	50–54	175	8.18
	55–59	158	7.39
	60–64	172	8.04
	65–69	91	4.25
	70–74	59	2.76
	75+	45	2.10
Sex	Male	529	24.73
	Female	1610	75.27
Race	Indigenous	104	4.86
	Gypsy (Rom)	3	0.14
	Raizal from the archipelago of San Andrés, Providencia, and Santa Catalina	3	0.14
	Black, mulatto (Afro-descendant), and Afro-Colombian	683	31.93
	None of the above	1346	62.93
Marital status	Married	178	8.32
	Separated or divorced	227	10.61
	Single	1221	57.08
	Common-law union	443	20.71
	Widowed	70	3.27
Level of educational	None	62	2.90
	Primary school (1st–5th levels)	577	26.98
	Secondary school (6th–9th levels)	384	17.95
	Media (10th–13th levels)	800	37.40
	Technical or technological (1–4)	237	11.08
	University (1–6)	77	3.60
	Postgraduate (1–4)
Occupation	Work	1073	50.16
	Household chores	581	27.16
	Job search	319	14.91
	Neither studying nor working, looking for work, or engaging in housework (NINIS)	76	3.55
	Study	60	2.81
	Permanently disabled	30	1.40
Years of residence in the neighborhood	10 years and older	452	21.13
	Between 5 and 10 years	350	16.36
	Between 1 and 5 years	821	38.38
	Less than 1 year	516	24.12

Note: * Age distribution follows the Departamento Administrativo Nacional de Estadística (DANE; Colombian Statistical Office) standard, which is used, for example, in population forecasts.

displays socioeconomic attributes grouped into seven standard variables. The database also includes two variables with open-text responses, namely, nationality and the relationship with the household head.

2.2. Perceptions of Security and Policy

Table 3. Perception of security at the city and neighborhood levels and anti-crime policy measures (n = 2139).

Variable	Category	N	(%)
From January to June 2024, have you heard if the Mayor’s Office or the police have implemented strategies to improve security and coexistence?	No	1592	0.74
	Yes	547	0.26
How would you rate the effectiveness of the strategies implemented by the Mayor’s Office or the police to improve security and coexistence? (n = 547) *	Very ineffective	355	0.65
	Ineffective	97	0.18
	Effective	88	0.16
	Very effective	7	0.01
On a scale of 1 to 5, where 1 is very unsafe and 5 is very safe, how would you describe your sense of security due to the presence of surveillance cameras in the city?	1. Very unsafe	343	0.16
	2. Unsafe	810	0.38
	3. Neither safe nor unsafe	789	0.37
	4. Safe	166	0.08
	5. Very safe	31	0.01
Compared with the situation 12 months ago (i.e., June 2023), what is the extent to which authorities in Cali have made changes to improve neighborhood security?	Is the same	1228	0.57
	Has increased	523	0.24
	Has decreased	220	0.10
	Does not know/Does not answer	168	0.08
On a scale of 1 to 5, where 1 is very unsafe and 5 is very safe, how safe do you feel in your neighborhood?	1. Very unsafe	336	0.16
	2. Unsafe	776	0.36
	3. Neither safe nor unsafe	664	0.31
	4. Safe	329	0.15
	5. Very safe	34	0.02
On a scale of 1 to 5, where 1 is very safe and 5 is very unsafe, how safe do you feel in Cali in general?	1. Very safe	120	0.06
	2. Safe	123	0.06
	3. Neither safe nor unsafe	781	0.37
	4. Unsafe	867	0.41
	5. Very unsafe	248	0.12

Note: * The first question is conditional on the second question.

presents six variables on perceptions of security and coexistence and policy related to these issues. Additionally, it incorporates two questions with multiple potential responses: “From January to June 2024 … (i) have you received information from the Mayor’s Office about:” and (ii) “which of these situations occurred in your neighborhood?”.

2.3. Sense of Security at Places

Figure 1 illustrates the variables associated with sense of security in nine private and public places.

2.4. Victimization, Sanction, and Self-Security Measures

Table 4. Victimization, security measures, and sanction probability perception (n = 2139).

Variable	Category	N	(%)
Apart from the incidents mentioned, in the last year, have you or a household member been a victim of any crime in Cali?	No	1409	0.66
	Yes	730	0.34
You consider that the likelihood of a crime being sanctioned in Cali is:	Very high	65	0.03
	High	93	0.04
	Medium	418	0.20
	Low	738	0.35
	Very low	730	0.34
	Does not know/answer	95	0.04

presents the variables for cases of victimization and probability of crime sanctions. Three complementary variables were captured using questions with multiple responses: “(i) In the last year, have you or a household member been a victim or experienced any of these? From January to June 2024 …. (ii) What measures did you take at home to protect yourself from crime? (iii) What measures did you take for your personal security?”

3. Methods

We conducted a pilot study one month prior to the survey. This pilot helped estimate the duration of the survey and provided valuable field feedback to improve the data collection instrument, further avoid open-ended questions, and incorporate multiple response pathways based on the participants’ responses. In addition, the direction of the Likert scale was reversed for certain questions to mitigate response acquiescence biases. This structure ensured a comprehensive dataset by preventing skipped responses.

Data were collected through in-person, non-probability convenience sampling in Santiago de Cali, Colombia, between 22 July and 16 August 2024, at SISBEN service points. The survey was conducted on respondents within these sites during fieldwork hours, who voluntarily agreed to participate. Therefore, inclusion in the sample depended on attendance at SISBEN service points and voluntary participation, instead of a probability-based selection mechanism. Using national identification, we linked the SISBEN IV survey data with the data collected by the current study—enabling the creation of the proposed integrated database. Notably, the matching success rate in this process was 100 percent, because all respondents were individuals registered in SISBEN IV, whether new or old.

The final dataset includes 2139 respondents. This value corresponds to 26.17% of the total population (N = 8172) based on the SISBEN-registered population in Cali during the same period (22 July–16 August 2024). The sample achieved should be interpreted as a result of convenience sampling at service points and not as a probabilistic sample from the full population. Consequently, conventional sampling error interpretations (e.g., margins of error derived from simple random sampling assumptions) are not strictly applicable.

To minimize missing data and field inconsistencies, the questionnaire was implemented on a digital platform with programmed restrictions (single-option constraints and automated skips based on prior responses). The participants were required to be of legal age (18+) and to provide informed consent, which is consistent with the Colombian law on data protection (Law 1266 of 2008). We omitted information on national IDs from the database to ensure full anonymity.

The sample is not fully representative of the population in Santiago de Cali. Participation was voluntary, and recruitment was conducted at SISBEN service points; this aspect may introduce self-selection that could overrepresent individuals with specific characteristics or extreme perceptions. For example, the dataset exhibits a notable gender imbalance (the sample is composed of approximately 75.3 per cent of women; Table 2), which may be related to patterns in program participation and the timing of data collection during standard weekday business hours. In general, women are the primary effective recipients of benefits from conditional cash transfer programs targeted by SISBEN. This trend is not exclusive to Colombia but is widely observed across Latin America [21]. Finally, the survey was conducted during regular working hours from Monday to Friday, which, given the lower participation rate of women in the labor force in Cali [22], increases the likelihood that women would attend offices where data collection was conducted.

This imbalance between the SISBEN IV-registered population and the survey dataset is also noted for other cardinal attributes (

Table 5. Mean-difference tests.

Variable	SISBEN IV-Registered Population (Mean)	Survey Dataset (Mean)	Difference (Population−Survey)
Age (years)	35.387	42.223	−6.836 **
Education (years)	9.680	8.396	1.283 **
Female share	0.668	0.753	−0.084 **
Single share	0.630	0.571	0.059 **
Obs.	8172	2139

Note: Computations are based on SISBEN IV and the dataset survey. Statistical significance ** 1%.

; regarding years of education, we constructed this variable using level of education and final grade). We strongly recommend considering these data characteristics when generalizing the findings apart from the population registered in SISBEN or those attending service points. Additionally, we observe that non-random survey participation influenced the probability of appearing in the survey [23]. Equation (1) presents a Probit selection model that contrasts such hypotheses in the current case:

$$\begin{array}{cc}\hfill {Survey}_i^{*}& ={\beta }_1+{\beta }_{2}Ag{e}_i+{\beta }_{3}Educatio{n}_i+{\beta }_{3}Wome{n}_i+{\beta }_{3}Singl{e}_i\hfill \\ & \hspace{1em} +u_i; u_i~N\left(0,1\right)\hfill \\ \hfill {Survey}_i^{*}& = 1\left[{Survey}_i^{*}> 0\right].\hfill \end{array}$$

(1)

Table 6. Results of the selection Probit.

Variable	Coefficient	Marginal Effect (AME)
Age	0.014708 **  [0.001162]	0.004560 **  [0.000352]
Education	−0.028076 **  [0.004320]	−0.008705 **  [0.001329]
Women	0.209339 *  [0.033545]	0.063450 **  [0.009924]
Single	−0.010137 *  [0.032495]	−0.003143 **  [0.010075]
Model fit	N = 8172	LogLik = −4493.269

Note: Computations are based on SISBEN IV and the dataset survey. Statistical significance ** 1%, * 5%. Standard errors are enclosed in square brackets [ ].

presents the estimation used in Equation (1). The model reveals that an increase in one age year increases the probability of appearing in the survey at 0.0045 percentage points (pp; column 3). Moreover, this same probability decreases by 0.003 pp if the individual is single and by 0.0087 pp for every additional year of education. Finally, being a woman increases the likelihood of appearing in the survey by 0.06 pp.

4. User Notes

Multiple analyses using this database are possible (e.g., for calculating indicators such as victimization rate (VR) [2]):

$$VR={\displaystyle \frac{T}{P }}$$

(2)

where T is the total number of victims of any crime, and P is the population aged 18 years and above. The numerator of VR is a response to the question: “In the last year, have you or a household member been a victim or experienced any of these? (Responses: theft or attempted theft in your home; theft or attempted theft of vehicles, vehicle parts, bicycles, motorcycles, or skateboards; deliberate destruction or damage to a house, store, or any other property that you or a member of your household owns; None; and Other.” Moreover, the denominator is different from DANE (15 years old), because the respondents were only adults aged 18 years old and above.

The database enables the presentation of spatial rate victimization using aggregate comunas [24,25] 3.

Table 7. Victimization by aggregate comunas (n = 2139).

Zone	Non-Victim		Victim		Total
	N	%	N	%
East	691	63.80%	392	36.20%	1083
Hillside	244	69.71%	106	30.29%	350
Central-North	192	58.90%	134	41.10%	326
Central-East	200	78.74%	54	21.26%	254
Rural Zone	56	64.37%	31	35.63%	87
North–South	26	66.67%	13	33.33%	39
Total	1409	65.87%	730	34.13%	2139

presents the results of the VR.

Alternatively, and given the potential self-selection biases discussed in the previous section, the study applied the partial identification framework proposed by Manski. Let us = P(S = 1) denote the probability of being observed in the survey, and p_s = E[Y|S = 1] the observed VR among the surveyed individuals by sex [23].

Table 8. Manski bounds (victimization): overall and by sex.

Metric	Value
s = P(survey = 1)	0.261747
p_s = E[Y|survey = 1]	0.341281
Lower = p_s ⁎ s	0.089329
Upper = p_s ⁎ s + (1 − s)	0.827582
p_s men (within survey)	0.302457
Bounds men [L,U]	[0.079167, 0.817420]
p_s women (within survey)	0.354037
Bounds women [L,U]	[0.092668, 0.830921]

Note: Computations are based on SISBEN IV and the dataset survey.

(with s = 0.26 [survey share] and p_s = 0.34 [victimization identified within the survey]) indicates that the Manski lower bound for the population VR is p_s ⁎ s = 0.089329, while the upper bound is p_s ⁎ s + (1 − s) = 0.827582. These bounds represent worst case scenarios and do not rely on ignorability assumptions. Disaggregating by sex, the observed VR among the surveyed men is 0.302, yielding bounds [0.079167, 0.817420]. For women, the observed rate is 0.354037, producing bounds [0.092668, 0.830921]. The higher lower bound for women denotes a higher observed VR in the survey sample, while wide intervals highlight the weak identification inherent to the absence of assumptions regarding non-surveyed individuals.

Furthermore, we estimate the average treatment effect on the treated (ATT) using propensity score matching (PSM), in which treatment is defined as being female. The propensity score is estimated once via the Probit model (Equation (1)).

Table 9. Average treatment effect on the treated by sex.

Neighbors (k)	ATT
1	0.006234
2	0.025873
3	0.027951
4	0.045615
5	0.042997

Note: Computations are based on SISBEN IV and the dataset survey 4.

reports ATT values estimated via PSM while varying the number of nearest neighbors (k). We keep caliper = 0.05 and impose common support. The propensity score is estimated once using the Probit model: treat (women = 1) ~ age + education + single. Sensitivity analysis was conducted by varying the number of nearest neighbors (k). The estimated ATT values are k = 1 (ATT = 0.0062), k = 2 (ATT = 0.0258), k = 3 (ATT = 0.0279), k = 4 (ATT = 0.0456), and k = 5 (ATT = 0.0429). These results indicate a consistently positive ATT, which suggests that, based on observables and within the survey, women experience a higher probability of victimization than comparable men.

Although PSM provides a conditional causal interpretation under the assumption of selection of observables, the Manski bounds serve as a complementary benchmark that remains valid under arbitrary selection. Taken together, these approaches bracket the potential range of population effects while preserving all estimated models and results ([23,26,27,28,29]).