Victimisation and fear of crime in Brazil: The effect on the desire to move

Abstract

We examined the effect of victimisation and fear of crime in Brazil and its attendant influence on the desire to move. Data drawn from the 2012 National Victimisation Survey were used to model the relationship between victimisation and the desire to move, with fear of crime used as a mediator variable. Our results indicate that being a victim of crime leads to increased fear, which, in turn, increases the desire to move. However, the indirect effect is comparatively small (i.e., our mediator variable—fear of crime—only marginally attenuates the relationship between victimisation and the desire to move). We also found evidence that property victimisation, victimisation occurring close to an individual's home, and indirect victimisation (i.e., hearing about a crime committed against another person in the neighbourhood) were also predictors of an increased desire to move in Brazil. Lastly, our results also suggest the importance of distinguishing between victimisation that occurred in the last 12 months compared to victimisation that occurred more than 12 months ago. Thus, policies that reduce victimisation and fear of crime may minimise the desire to move and any related economic and social costs.

Keywords

Brazil fear of crime mediation moving intention victimisation

Introduction

Although there are signs of decline in the 2017–2018 period, Brazilian homicide rates are continuously ranked among the highest violent crime rates in the world. Recent data from the World Bank reveals that, in 2018, almost 60,000 people were murdered in Brazil, with 27.4 victims per 100,000 inhabitants. Only El Salvador (52.0), Jamaica (43.9), Honduras (38.9), Venezuela (36.7), and Mexico (29.1) had higher homicide rates per 100,000 inhabitants. In contrast, the United States, for example, was in the 28^th position, with a rate of five victims per 100,000 inhabitants (16,214 people were murdered in the US in 2018).

However, homicide is not the only concern in Brazil. Other crimes, such as robbery, theft, drug trafficking, and aggression, also present at high rates (Fórum Brasileiro de Segurança Pública, 2021). This helps to explain why the Brazilian population reported the highest rates of fear of violent crime in the world in 2021 (Global Peace Index, 2022). Against this background, several adverse consequences may be observed in victims’ daily routines. Precautionary behaviour such as avoiding risky places or people, not walking in the dark or using public transport, and staying at home more often are some examples of the consequences of crime (e.g., Averdijk, 2011; Garofalo, 1981; Skogan & Maxfield, 1982; Rountree & Land, 1996).

Researchers have also shown that violence influences the desire to move, and these studies can be generally divided into two strands. First, from the macro-level perspective, it is possible to see how violence in a region increases people's desire to migrate to other cities, states, and countries (e.g., Boggess & Hipp, 2010; Liska & Bellair, 1995; Roth et al., 2020). Second, micro-level analysis allows one to understand how personal victimisation may impact the decision to move (e.g., Dugan, 1999; McNeeley & Stutzenberger, 2013; Xie & McDowall, 2008b). It is important to note that personal victimisation can also include indirect victimisation, where individuals witness the effects of crimes suffered by others (Roth et al., 2020; Xie & McDowall, 2008b).

Decisions to move may generate substantial monetary and psychological costs for victims (Najdowski & Ullman, 2009; Yetter, 2018). The related financial expenses comprise direct costs associated with breaking rental agreements and indirect costs like missing work to search for a new residence. The psychological costs of moving can also result in the victim experiencing prolonged emotional and social distress (Dugan, 1999).

Against this background, we use data from the 2012 Brazilian National Victimisation Survey (NVS) to examine whether victimisation is associated with an increased desire to move. We also explored how this association was influenced by: (i) the type of victimisation (e.g., person or property), (ii) the location of victimisation (e.g., at home or elsewhere), (iii) the proximity of victimisation (e.g., your next-door neighbour was the victim), and (iv) when the victimisation occurred (e.g., within 12 months).

Moreover, studies have identified that fear of crime mediates the relationship between victimisation and the desire to move once people believe their house and neighbourhood are less safe (Engel & Ibáñez, 2007; McNeeley & Stutzenberger, 2013; Roth et al., 2020). Therefore, we used mediation analysis with fear of crime at the individual and neighbourhood levels as the mediator variable. To examine and test these relationships, we employ the recent parametric regression-based causal mediation analysis developed by Samoilenko and Lefebvre (2021).

The remainder of the article is structured as follows. Section 2 includes an overview of the salient theoretical and empirical literature. Section 3 reviews the dataset and empirical strategy. The results are presented and discussed in section 4. The final section presents concluding remarks.

Theoretical and empirical review

Crime profoundly affects society in several ways. The economic cost of crime is substantial, impacting society financially and emotionally (Fadaei-Tehrani & Green, 2002). Victimisation from criminal acts may generate a constant feeling of fear that influences individual choices. It may also increase the perception that new victimisation is likely, and, hence, influence the desire to move (McNeeley & Stutzenberger, 2013; Xie & McDowall, 2008b).

The neighbourhood also plays an essential role in individual well-being, as personal safety and security are often considered important neighbourhood characteristics (Leibbrand et al., 2021). Moreover, high levels of neighbourhood violence may adversely affect property values and encourage local retail and service providers to leave the area (Hipp et al., 2009).

In order to understand the determinants of moving, researchers have analysed factors at the macro-level and micro-level. For example, in a macro-level study, Cullen and Levitt (1999) investigated the relationship between rising city crime rates and urban flight. According to the authors, a rising crime rate reduces the city population due to increased out-migration. The authors also observed that households with higher levels of education or children present are more responsive to changes in crime rates (Cullen & Levitt, 1999).

These macro-level studies may be augmented by micro-level evidence. Micro-level studies examine the reasons for mobility at the individual level rather than the aggregate movement of people across geographical boundaries (Xie & McDowall, 2008b). In a micro-level study, Dugan (1999) tested the hypothesis that criminal victimisation is associated with an increased likelihood of moving. The findings showed that property crime victimisation (e.g., burglaries and car thefts) might influence moving decisions more than violent victimisation (e.g., rape, robbery, and assault). One potential explanation is that violent victimisation tends to be committed by a person known to the victim, while property victimisation is more likely to be committed by strangers (McNeeley & Stutzenberger, 2013).

The effect of violence can also be divided into two groups: direct and indirect victimisation. Direct victimisation involves the household, while indirect victimisation involves violence experienced by others (Xie & McDowall, 2008b). Several authors have shown that direct experience of crime near an individual's home affects the desire to move (Dugan, 1999; McNeeley & Stutzenberger, 2013; Xie & McDowall, 2008a; Yetter, 2018). The literature also provides evidence of indirect victimisation, or the knowledge of crime experienced by friends, neighbours, or family members on residential mobility (Xie & McDowall, 2008b). Indirect victimisation may also heighten the perception of residing in an unsafe neighbourhood.

According to McNeeley and Stutzenberger (2013), the perception of neighbourhood safety can mediate the effect of property victimisation on the desire to change residence. In other words, those who have experienced crime near home tend to move, assuming their home is less safe (McNeeley & Stutzenberger, 2013). The authors also contend that adding risk perception to the model may aid in understanding the effect of indirect victimisation. It may happen because experience with crime or victimisation increases the perception of risk and the feeling of being unsafe, which changes the routine regarding safety precautions (Rountree & Land, 1996). The change in the routine may extend to other ways to increase safety, including moving to a safer neighbourhood.

Moving may serve as a powerful mechanism for reducing exposure to violence (Leibbrand et al., 2021; McNeeley & Stutzenberger, 2013). On this point, empirical studies have found a positive relationship between experiencing (direct or indirect) victimisation and the desire to move (Dugan, 1999; Xie & McDowall, 2008b). It is important to note that the effect of property victimisation near the home is more important than personal victimisation (Dugan, 1999; Xie & McDowall, 2008b). The decision to move from a violent neighbourhood is a way to enhance long-term well-being, even if it generates some sacrifices, such as having less access to family, friends, or other resources (Leibbrand et al., 2021).

Against this background, the present study considers important questions already discussed by the recent research on the individual-level analysis of victimisation on the desire to move. We include in our analysis different types of victimisation, as discussed by Dugan (1999) and McNeeley and Stutzenberger (2013) and indirect victimisation, as examined by Xie and McDowall (2008b). Furthermore, we extend existing research by considering whether: (i) victimisation occurred within the last 12 months or (ii) victimisation occurred more than 12 months ago. Moreover, unlike the previous studies that employ data from the US or Central America (e.g., Dugan, 1999; McNeeley & Stutzenberger, 2013; Roth et al., 2020; Wood et al., 2010), we investigate these relationships using a large and representative dataset from Brazil.

Several papers have analysed the effect of victimisation in moving decisions over different periods. For example, some studies considered the effects of victimisation in the last six months (Dugan, 1999; Xie & McDowall, 2008b, 2014), within the last 12 months (Ferraro, 1997; Roth et al., 2020), and within the last 24 months (McNeeley & Stutzenberger, 2013). However, there is no evidence comparing the impact of moving intentions over different periods (e.g., victimisation that occurred within the last 12 months compared to victimisation that occurred more than 12 months ago). Thus, we contribute to the literature by comparing the effects of victimisation within the last 12 months to victimisation that occurred more than 12 months ago on the desire to move.

Methodology

Data and variables

To examine these relationships, we used NVS data collected in 2012 by the Brazilian Ministry of Justice through the National Secretariat for Public Security (SENASP, 2012). We employed individuals as the unit of analysis, and the sample consisted of 61,787 observations from all 26 Brazilian state and federal districts. The sample selected individuals older than 16, residing in municipalities with more than 15,000 inhabitants in the major urban cities of Brazil (346 municipalities). The survey contained data on victimisation, fear of crime, and desire to move as well as information on the respondents’ demographic, economic, and social characteristics.

The dependent variable was “desire to move”. The specific question was, “If you could choose, would you like to continue living in your residence?”. If individuals answered “yes”, we recoded the dependent variable as “0”; if the answer was “no”, the variable was recoded as “1”.

Rather than measuring whether an individual moved, our study measures whether an individual intended to move. Victimisation related to mobility histories is typically obtained through longitudinal data (e.g., Dugan, 1999; Xie & McDowall, 2008b, 2014). However, due to the difficulty and the high cost of compiling this type of data (Baydar & White, 1988), other studies have relied on cross-sectional data and focused on the desire to move (e.g., McNeeley & Stutzenberger, 2013; Roth et al., 2020; Skogan 1992). On this point, data from the United States suggest that around 20% of people who reported a desire to move actually did move the following year (Mateyka, 2015).

Regarding our independent variables, we use two groups of victimisation variables: (i) personal victimisation (i.e., aggression and threat)¹ and (ii) property victimisation (i.e., robbery and theft).² In other words, four crime types are specified to determine whether the effects of victimisation on the desire to move varied. Studies typically find that property victimisation is positively related to moving out of a neighbourhood (Dugan, 1999; McNeeley & Stutzenberger, 2013; Xie & McDowall, 2008b). In examining these relationships for Brazil, we can shed light on how personal and property victimisation influence the desire to move.

Specifically for robbery and property crime victimisation, we use available information on where the crime has occurred (the location of other crime types is unavailable). If the answer was “the robbery took place in my house”, we compare it with the effect of the robbery occurring in another place.

It is also important to note that the decision or the desire to move may be influenced when a crime occurs near or at home (Dugan, 1999; Xie & McDowall, 2008b). We also examined if indirect victimisation (i.e., crime in the neighbourhood area) influenced the desire to move. We considered two different kinds of indirect victimisation: (i) respondents hearing about other people being physically assaulted (i.e., aggression) and (ii) respondents hearing about other people being robbed. It is worth noting that the survey does not identify if the respondent knew the victim. As a result, we only know that a respondent heard about a crime committed against another person in the neighbourhood. However, as direct victimisation is rare (Warr, 1994), our analysis can provide a greater understanding of the effect of indirect victimisation on the desire to move.

Another interesting aspect is the different victimisation reporting timelines found in the literature. Some authors used victimisation in the last six months (e.g., Dugan, 1999; Xie & McDowall, 2008b, 2014), in the last 12 months (Ferraro, 1997; Roth et al., 2020), and others in the last 24 months (e.g., McNeeley & Stutzenberger, 2013). On this point, the data in the NVS provide us with a unique opportunity to compare the effects of property victimisation (robbery and theft) within the last 12 months to property victimisation that occurred more than 12 months ago on the desire to move.

The literature has also noted that fear of crime operates as a channel between victimisation and the desire to move (McNeeley & Stutzenberger, 2013; Yetter, 2018). Subsequently, we use “fear at home” and “fear in the neighbourhood” as mediator variables. The questions were: “How do you feel when you are alone at home?” and “How do you feel walking on the streets of the neighbourhood where you reside?” If respondents who answered those questions were “very insecure” or “a little insecure”, we classified them as having a fear of crime (i.e., 1 = fear of crime; 0 = otherwise).

We also included variables to control for (i) demographic and economic factors, (ii) domicile and municipality characteristics, and (iii) neighbourhood characteristics (Table 1). Regarding demographic and economic factors, we included controls for gender, age, income, and so on. Individuals were also classified according to their level of educational attainment (no education, middle school, high school, or more than high school) and income (less than R$678, from R$678.01 to R$1,356, from R$1,356.01 to R$2,034, from R$2,034.01 to R$2,712, or more than R$2,712). The level of education and income may positively influence an individual's desire to move.

Table 1.

Definitions and summary statistics.

Variables	Description	Mean
Dependent Variable: Desire to move
Desire to move	1 = Desire to move; 0 = otherwise	0.678
Independent Variables: Victimisation
Theft	1 = Claimed to be a victim of theft in the last year; 0 = otherwise	0.110
Robbery	1 = Claimed to be a victim of a robbery in the last year; 0 = otherwise	0.033
Threat	1 = Claimed to be a victim of threat in the last year; 0 = otherwise	0.126
Aggression	1 = Claimed to be a victim of aggression in the last year; 0 = otherwise	0.024
Mediate Variables: Fear of Crime**
Fear at home	1 = If the respondent claimed insecurity at home; 0 = otherwise	0.154
Fear in the neighbourhood	1 = If the respondent claimed insecurity in the neighbourhood; 0 = otherwise	0.457
Demographic and economic factors
Male	1 = Male, 0 = female	0.436
Unemployed	1 = Unemployed, 0 = otherwise	0.041
16–24 years old	1 = 16 to 24 years old; 0 = otherwise	0.157
25–34 years old	1 = 25 to 34 years old; 0 = otherwise	0.206
35–44 years old	1 = 35 to 44 years old; 0 = otherwise	0.19
45–59 years old	1 = 45 to 59 years old; 0 = otherwise	0.242
More than 59 years old (Ref)	1 = More than 59 years old; 0 = otherwise (Ref.)	0.205
Married	1 = Married; 0 = otherwise	0.575
Non-white	1 = Non-white; 0 = otherwise	0.536
No education	1 = No education; 0 = otherwise	0.069
Middle school	1 = Studied until middle school; 0 = otherwise	0.422
High school	1 = Studied until high school; 0 = otherwise	0.365
More than high school (ref)	1 = Studied more than high school; 0 = otherwise	0.144
Up to R$678*	1 = Receive income up o R$678; 0 = otherwise	0.218
From R$678.01 to R$1,356	1 = Receive income from R$678,01 to 1356; 0 = otherwise	0.314
From R$1,356.01 to R$2,034	1 = Receive income from R$1356.01 to 2034; 0 = otherwise	0.198
From R$2,034 to R$2,712	1 = Receive income from R$2034.01 to 2712; 0 = otherwise	0.161
More than R$2,712 (Ref)	1 = Receive income up o R$2712; 0 = Otherwise (Ref.)	0.109
Domicile and Municipality characteristics
People in home	Number of people residing in the home	3.45
Children proportion	The proportion of children in the home	0.181
Years at current neighbourhood	Number of years living in the neighbourhood	17.09
Apartment	1 = Lives in an apartment; 0 = otherwise	0.08
House	1 = Lives in a house; 0 = otherwise	0.895
Low-cost housing (Ref)	1 = Lives in a low-cost housing; 0 = otherwise (Ref.)	0.025
Capital	1 = Lives in a capital of a state; 0 = otherwise	0.221
Capital nearby	1 = Lives in a municipality that has contiguity with a capital; 0 = otherwise	0.151
Countryside (Ref)	1 = Lives in a countryside; 0 = otherwise (Ref.)	0.628
Mega cities	1 = Resides in a city of over 500,000 people; 0 = otherwise	0.258
Large cities	1 = Resides in a city from 300,000 to 500,000 people	0.207
Medium cities	1 = Resides in a city from 50,000 to 300,00 people; 0 = otherwise	0.42
Small cities (Ref)	1 = Resides in a city under 50,000 people; 0 = otherwise (Ref.)	0.115
Southeast	1 = Resides in Southeast of Brazil; 0 = otherwise	0.372
South	1 = Resides in South of Brazil; 0 = otherwise	0.156
Northeast	1 = Resides in Northeast of Brazil; 0 = otherwise	0.283
Central-West	1 = Resides in Central-West of Brazil; 0 = otherwise	0.084
North (Ref)	1 = Resides in North of Brazil; 0 = otherwise (Ref.)	0.105
Neighbourhood characteristics
Abandoned places	1 = There are abandoned places in the neighbourhood; 0 = otherwise	0.161
Abandoned cars	1 = There are abandoned cars in the neighbourhood; 0 = otherwise	0.064
Empty lot	1 = There is an empty lot in the neighbourhood; 0 = otherwise	0.371
Gunshots being fired	1 = Gunshots are being fired in the neighbourhood; 0 = otherwise	0.287
Unpleasant odours	1 = There are unpleasant odors in the neighbourhood; 0 = otherwise	0.345
Loud noise	1 = There are loud noises in the neighbourhood; 0 = otherwise	0.426

*The minimum wage in 2013 was R$678 per month, representing about US$ 335,65.

**We also created a composite fear dummy variable (1 = the individual reported “fear at home” and “fear in the neighbourhood”; 0 = otherwise). Our subsequent empirical results are virtually identical if we use this composite measure of fear.

We also included control variables to account for different household and municipality characteristics (e.g., the number of people in the household, proportion of children, years in the current neighbourhood, house and city type, city size, and Brazilian region). For municipal characteristics, we considered if the locality was a capital city, located near a capital city, or classified as a regional area. On this point, we also controlled for the population size of the city: small (<50,000 people), medium (between 50,000 and 300,000 people), large (between 300,000 to 500,000 people), and mega (>500,000 people).

Lastly, we included control variables to account for commonly reported neighbourhood disorders. More specifically, we controlled for the reported presence of abandoned homes and cars, gunshots, and loud sounds. Evidence suggests that disorderly neighbourhood conditions affect victimisation and an individual's risk perception (McNeeley & Stutzenberger, 2013; Tita et al., 2006).

Mediation analysis: Victimisation, fear of crime, and the desire to move

As noted above, our dependent variable is the desire to move (1 = yes; 0 = no). Victimisation—our primary independent variable—may influence the desire to move directly or indirectly via fear of crime. Our victimisation and fear of crime variables are coded as binary variables. Therefore, we employ a mediation model where our outcome variable is the desire to move, our treatment variable is victimisation, and our mediator variable is fear of crime.

Logistic regression is commonly used to perform a conventional mediation analysis (Samoilenko & Lefebvre, 2021). These studies employ the approximate mediation approach of Valeri and VanderWeele (2013), in which the rare outcome assumption (ROA) is essential (Lai et al., 2020; Shreffler et al., 2020). The dependent variable analysed should be relatively rare (Valeri & VanderWeele, 2013). Otherwise, the approximate mediation approach will overestimate the direct effect and underestimate the indirect effect (Sheikh et al., 2017), and the results will be biased if logistic regression is used (Valeri & VanderWeele, 2013).³

Some strategies have been used to circumvent the ROA assumption. For example, Valeri and VanderWeele (2013) also presented the log-linear model when the outcome is not rare, which has already been tested by the literature (e.g., Rø et al., 2017). However, some problems related to convergence were found by using that strategy (Nguyen et al., 2016).

Therefore, some authors (e.g., Gaynor et al., 2019; Samoilenko et al., 2018) propose methods that are not subjected to the ROA condition and convergence problem. The most recent model developed by Samoilenko and Lefebvre (2021) is the “exact mediation approach”, which differs from the approximate estimator of Valeri and VanderWeele (2013). The authors found that when the outcome was common, the bias and variance of the exact estimator were systematically smaller than those of the popular approximate estimator (Samoilenko & Lefebvre, 2021).

Since our outcome—the desire to move—is not rare (approximately 32% of our sample have expressed a desire to move), we applied the exact mediation approach to finding the direct (i.e., natural direct effect or NDE) and indirect (i.e., natural indirect effect or NIE) effects of victimisation on the desire to move. Therefore, we employ, following Samoilenko and Lefebvre (2021), the logistic regression models below:

L o g i t (P (M = 1 | A = a, C = c)) = β_{0} + β_{1} a + β_{2}^{^{'}} c

(1)

L o g i t (P (Y = 1 | A = a, M = m, C = c)) = θ_{0} + θ_{1} a + θ_{2} m + θ_{3}^{^{'}} c

(2)

where Y is the binary outcome (desire to move), M is the binary mediator (fear of crime), A is the binary exposure (victimisation), and C is the set of control variables (related to demographic factors, domicile characteristics, and municipality characteristics). Using equations (1) and (2), it is possible to present explicit expressions for the direct (NDE) and indirect (NIE) effects as odds ratios:

O R_{a, a^{*} | c}^{N D E} = \frac{\frac{P (Y (a, M (a^{*})) = 1 | C = c)}{1 - P (Y (a, M (a^{*})) = 1 | C = c)}}{\frac{P (Y (a^{*}, M (a^{*})) = 1 | C = c)}{1 - P (Y (a^{*}, M (a^{*})) = 1 | C = c)}}

(3)

O R_{a, a^{*} | c}^{N I E} = \frac{\frac{P (Y (a, M (a)) = 1 | C = c)}{1 - P (Y (a, M (a)) = 1 | C = c)}}{\frac{P (Y (a, M (a^{*})) = 1 | C = c)}{1 - P (Y (a, M (a^{*})) = 1 | C = c)}}

(4)

Note that equation (3), related to NDE, compares

Y (a, M (a^{*}))

Y (a^{*}, M (a^{*}))

, where the aim is to find the probability of Y assuming a value of 1 when the change in the exposure level to victimisation from

A = a^{*}

A = a

is observed. Equation (4) is defined as a difference between

Y (a, M (a))

with

Y (a, M (a^{*}))

. That is, the probability of Y assuming a value of 1 when the mediator variable changes its status in response to a change in victimisation status. The interaction between NDE and NIE results in the total effect (TE):

O R_{a, a^{*} | c}^{T E} = O R_{a, a^{*} | c}^{N D E} \times O R_{a, a^{*} | c}^{N I E}

(5)

To find these three effects related to victimisation, fear of crime, and desire to move, we consider four different crime exposure types (i.e., theft, robbery, threat, and aggression) and two different fear of crime types (i.e., at home and in the neighbourhood). These overarching relationships are summarised in Figure 1.

Figure 1.

The relationship between victimisation (A), fear of crime (M), and desire to move (Y).

More specifically, Figure 1 shows that the TE of victimisation on the desire to move may be partially mediated by fear of crime (paths A, M, and Y). A significant natural indirect effect of the mediated pathway would support that hypothesis. It is important to note that four different models related to the desire to move are presented: (1) effects of different types of victimisation (i.e., threat, aggression, theft, and robbery); (2) victimisation location effects (i.e., at home versus elsewhere); (3) indirect victimisation effects (when a neighbour was victimised); and (4) victimisation time frame effects (within 12 months versus more than 12 months). These four models use “fear at home” and “fear in the neighbourhood” as mediator variables.

Results and discussion: Fear of crime and desire to move

Table 1 presents the definitions and summary statistics for the variables used in our mediation analysis. In our sample, 67.80% declared wanting to move from home. In contrast, McNeeley and Stutzenberger (2013) find much lower values for the US (approximately 30.00% stated a desire to move from their homes). Our independent variable—victimisation—is divided into four categories (i.e., theft, robbery, threat, and aggression). In Table 1, threat (12.60%) and theft (11.00%) are the more common victimisations. Our mediator variables are “fear at home” and “fear in the neighbourhood”, and a relatively large proportion of our sample reported feeling insecure in their neighbourhood (45.70%).

Regarding our demographic variables, 43.60% of our samples are men. Also, only a relatively small proportion of our sample is unemployed (4.10%). In terms of age, which is divided into five categories, the largest category comprises individuals 45–59 (24.20%). Most of our sample reported being married (57.50%) and nonwhite (53.60%). Turning to education, only 14.40% of our sample reported having more than a high school diploma. Concerning income, more than 73% of our sample received up to three times the minimum wage (up to R$ 1,530). Similar interpretations can be made for other variables.

In Table 2, we report the results from our logistic regression. The results indicate that respondents who reported being victimised had higher levels of fear (at home and in the neighbourhood) and a stronger desire to move. More specifically, respondents who reported being a victim of theft were 35.40% more likely to be fearful at home. It is worth noting that being a victim of theft increased the desire to move by 15.60% while being fearful at home also increased the desire to move by 33.33%. However, the relationship between victimisation and fear may be overstated, hence, the importance of using mediation analysis (Figure 1).

Table 2.

Logistic regression testing the effects of victimisation and fear of crime on the desire to move.

Variables	Fear at home (1)		Desire to move (1)		Fear at neighbourhood (2)		Desire to move (2)
	Coef.	SE.	Coef.	SE.	Coef.	SE.	Coef.	SE.
	Victimisation
Theft	1.354***	(0.047)	1.156***	(0.033)	1.352***	(0.038)	1.146***	(0.033)
Robbery	1.311***	(0.077)	0.988	(0.049)	1.511***	(0.077)	0.971	(0.048)
Threat	1.308***	(0.045)	1.645***	(0.046)	1.283***	(0.036)	1.635***	(0.046)
Aggression	1.089	(0.077)	1.086	(0.065)	1.028	(0.063)	1.089	(0.066)
Fear of Crime
Fear at home			1.333***	(0.033)
Fear in the neighbourhood							1.409***	(0.027)
Demographic and Economic
Male	0.360***	(0.010)	0.881***	(0.017)	0.570***	(0.010)	0.888***	(0.017)
Unemployed	0.961	(0.056)	1.202***	(0.054)	1.011	(0.044)	1.199***	(0.054)
16–24 years old	1.259***	(0.061)	1.357***	(0.052)	0.953	(0.034)	1.373***	(0.052)
25–34 years old	1.228***	(0.056)	1.282***	(0.045)	0.899***	(0.029)	1.302***	(0.046)
35–44 years old	1.050	(0.046)	1.317***	(0.045)	0.929**	(0.029)	1.326***	(0.045)
45–59 years old	1.073*	(0.042)	1.259***	(0.038)	0.964	(0.026)	1.265***	(0.038)
Married	1.109***	(0.028)	0.973	(0.020)	1.081***	(0.021)	0.971	(0.020)
Non-white	0.962	(0.025)	1.097***	(0.022)	0.931***	(0.018)	1.101***	(0.022)
No education	1.353***	(0.087)	0.876***	(0.044)	1.030	(0.048)	0.882**	(0.044)
Middle school	1.313***	(0.059)	0.829***	(0.028)	1.110***	(0.035)	0.830***	(0.028)
High school	1.130***	(0.048)	0.912***	(0.028)	1.084***	(0.031)	0.909***	(0.028)
Up to R$678	1.374***	(0.078)	1.290***	(0.053)	0.880***	(0.034)	1.317***	(0.055)
From R$678.01 to R$1,356	1.384***	(0.073)	1.162***	(0.043)	0.943*	(0.033)	1.179***	(0.044)
From R$1,356.01 to R$2,034	1.258***	(0.067)	1.122***	(0.042)	0.949	(0.033)	1.134***	(0.043)
From R$2,034 to R$2,712	1.239***	(0.067)	1.077**	(0.040)	0.933**	(0.032)	1.089**	(0.041)
Domicile and Municipality
People in home	1.008	(0.008)	0.988*	(0.007)	0.995	(0.006)	0.989*	(0.007)
Children proportion	0.921	(0.062)	1.384***	(0.074)	1.068	(0.055)	1.374***	(0.074)
Years at current neighbourhood	0.997***	(0.001)	0.985***	(0.001)	0.995***	(0.001)	0.985***	(0.001)
Apartment	0.553***	(0.048)	0.794***	(0.052)	1.050	(0.068)	0.776***	(0.050)
House	0.938	(0.064)	0.689***	(0.038)	0.911*	(0.051)	0.692***	(0.038)
Capital	1.059	(0.049)	1.108***	(0.041)	0.985	(0.035)	1.113***	(0.041)
Capital nearby	1.177***	(0.042)	1.217***	(0.034)	1.157***	(0.031)	1.214***	(0.034)
Mega cities	1.243***	(0.070)	1.358***	(0.059)	2.423***	(0.102)	1.281***	(0.056)
Large cities	1.148***	(0.052)	1.123***	(0.040)	2.070***	(0.069)	1.068*	(0.038)
Medium cities	1.036	(0.041)	1.094***	(0.034)	1.492***	(0.044)	1.064**	(0.034)
South	0.400***	(0.020)	0.957	(0.039)	0.415***	(0.016)	0.987	(0.040)
Southeast	0.474***	(0.020)	1.381***	(0.049)	0.427***	(0.014)	1.431***	(0.051)
Northeast	0.581***	(0.022)	1.186***	(0.040)	0.720***	(0.023)	1.187***	(0.040)
Central-West	0.565***	(0.029)	0.953	(0.042)	0.663***	(0.027)	0.960	(0.042)
Neighbourhood
Abandoned places	1.072**	(0.034)	1.087***	(0.028)	1.065**	(0.026)	1.085***	(0.028)
Abandoned cars	0.993	(0.047)	1.097**	(0.041)	1.035	(0.038)	1.093**	(0.040)
Empty lot	1.092***	(0.028)	1.075***	(0.022)	1.175***	(0.023)	1.065***	(0.022)
Gunshots being fired	1.348***	(0.037)	1.389***	(0.030)	1.688***	(0.036)	1.351***	(0.029)
Unpleasant odors	1.261***	(0.033)	1.379***	(0.029)	1.350***	(0.027)	1.361***	(0.029)
Loud noise	1.213***	(0.031)	1.460***	(0.029)	1.328***	(0.025)	1.441***	(0.029)
Estimated population	61,787		61,787		61,787		61,787
Pseudo R²	0.0764		0.0728		0.0852		0.0752
Wald	3,726.18		5,052.92		6,228.32		5,195.75

Note. Coefficients are presented in odds ratios. ***p < 0.01, **p < 0.05, *p < 0.10. For specification (1), the mediator variable is fear at home. For specification (2), the mediator variable is fear in the neighbourhood. For both specifications, treatment and outcome variables are victimisation (theft, robbery, threat, and aggression) and desire to move, respectively.

Looking at Table 2, other relationships are evident. For example, men tend to be less fearful at home and in the neighbourhood and, consequently, have less desire to move. We also note that compared to employed respondents, unemployed people tend to have a higher desire to move. Another interesting result is related to age. Our results show that older people present a lower probability of moving. Compared to nonmarried respondents, we find no evidence of a statistically significant association between marriage and the desire to move. Compared to white people, nonwhites are more likely to want to move. Concerning education, our model produced mixed results. For example, compared to those respondents with a high school education, respondents with a middle school education are more likely to be fearful but have a lower desire to move. Finally, compared to those respondents in the highest income bracket (i.e., >R$2,712), respondents in all four income categories have a higher desire to move.

Regarding our domicile and municipality characteristics, the number of years in the current neighbourhood tends to reduce the desire to move. Another interesting result relates to living in an apartment and fear at home. Living in an apartment reduces the likelihood of fear at home by 44.70%. We also note that the city's size significantly influences the fear and desire to move. For example, living in mega cities increases the likelihood of fear in the neighbourhood and the desire to move. People in the Southeast tend to have a greater desire to move. Abandoned places, empty lots, unpleasant odors, and loud noise increased the likelihood of fear and desire to move. However, the most influential characteristic is hearing gunshots being fired, which increased the fear of crime by 68.80% and the desire to move by 35.10%.

It is important to note that in Table 2, our model reveals only the direct effect of the desire to move. Based on our theoretical framework, we expect that victimisation leads to fear, which, in turn, increases the desire to move. As a result, we may overestimate the direct and indirect effects when these two variables are included in our regression analysis. Thus, to address this problem, our subsequent logistic regression analysis estimates the direct and indirect effects on the desire to move (with victimisation as the treatment variable and fear as a mediator variable).

In Table 3 (Model 1), we present the direct and indirect effects of victimisation on the desire to move, using fear as the mediator variable. For example, in the first column, with fear at home as a mediator variable, we see that theft victims are 16.90% more likely to want to move (15.60% direct and 1.10% indirect). For fear in the neighbourhood, our model shows a Total Effect (TE) of 17.40% (14.50% direct and 2.60% indirect). Regarding the other types of victimisation, we observe similar results. For example, exposure to threats increases the desire to move by more than 66%.

Table 3.

The direct and indirect effects of predicting the desire to move by victimisation type.

Variables	Theft		Robbery		Threat		Aggression
	Coef.	SE.	Coef.	SE.	Coef.	SE.	Coef.	SE.
	Mediate: Fear at home
Direct (NDE)	1.156***	(0.033)	0.988	(0.048)	1.644***	(0.046)	1.086	(0.064)
Indirect (NIE)	1.011***	(0.002)	1.010***	(0.002)	1.010***	(0.002)	1.003	(0.003)
Total effect	1.169***	(0.033)	0.998	(0.049)	1.660***	(0.046)	1.089	(0.065)
Mediate: Fear in the neighbourhood
Direct (NDE)	1.145***	(0.032)	0.971	(0.047)	1.630***	(0.045)	1.089	(0.064)
Indirect (NIE)	1.026***	(0.003)	1.036***	(0.005)	1.022***	(0.003)	1.002	(0.005)
Total effect	1.174***	(0.033)	1.006	(0.049)	1.665***	(0.046)	1.091	(0.065)
No. of victims	6,769		2,051		7,773		1,463

Note. Coefficients are presented as odds ratios. ***p < 0.01, **p < 0.05, *p < 0.10. All regression models include the control variables listed in Table 2. The sample is comprised of 61,787 individuals.

NDE: natural direct effect; NIE: natural indirect effect.

In Table 4, we present the results stratified by victimisation location (e.g., at home or elsewhere). The results were statistically significant only for the indirect relationship, and they show that the effect of victimisation tends to be higher when the crime happens near an individual's home. However, we could only test this hypothesis for robbery due to data limitations. In Table 5, we present the results of indirect victimisation. Looking at the results, we note that aggression directly affects an individual's desire to move (however, there is little, if any, evidence of an indirect effect). In contrast, however, robbery only affects the desire to move indirectly via its impacts of fear at home and in the neighbourhood.

Table 4.

The direct and indirect effects of predicting the desire to move by location of victimisation.

Variables	Robbery (home)		Robbery (other)
	Coef.	SE.	Coef.	SE.
	Mediate: Fear at home
Direct (NDE)	1.040	(0.120)	0.979	(0.052)
Indirect (NIE)	1.023***	(0.006)	1.008***	(0.003)
Total effect	1.064	(0.123)	0.986	(0.053)
Mediate: Fear in the neighbourhood
Direct (NDE)	1.033	(0.119)	0.960	(0.051)
Indirect (NIE)	1.047***	(0.011)	1.034***	(0.005)
Total effect	1.082	(0.125)	0.992	(0.053)
No. of victims	355		1,700

NDE: natural direct effect; NIE: natural indirect effect.

Table 5.

The direct and indirect effects of predicting the desire to move by indirect victimisation.

Variables	Aggression (Indirect victimisation)		Robbery (Indirect victimisation)
	Coef.	SE.	Coef.	SE.
	Mediate: Fear at home
Direct (NDE)	1.244***	(0.030)	1.007	(0.023)
Indirect (NIE)	1.001	(0.001)	1.009***	(0.001)
Total effect	1.245***	(0.030)	1.016	(0.023)
Mediate: Fear in the neighbourhood
Direct (NDE)	1.238***	(0.030)	0.967	(0.022)
Indirect (NIE)	1.007***	(0.002)	1.056***	(0.004)
Total effect	1.246***	(0.030)	1.021	(0.023)
No. of victims	13,256		19,892

NDE: natural direct effect; NIE: natural indirect effect.

Finally, in Table 6, we present our findings for those respondents who reported being victimised in the last 12 months compared to those who reported being victimised over 12 months ago. In reviewing Table 6, thefts in the last 12 months had a higher impact on the desire to move compared to thefts that happened over 12 months ago. For robberies in the last 12 months, we observed a statistically significant indirect effect on the desire to move. However, we observed statistically significant direct and indirect effects on the desire to move for robberies that occurred more than 12 months ago.

Table 6.

The direct and indirect effects of predicting the desire to move by victimisation time frame.

Variables	Theft (LTM)		Robbery (LTM)		Theft (ATM)		Robbery (ATM)
	Coef.	SE.	Coef.	SE.	Coef.	SE.	Coef.	SE.
	Mediate: Fear at home
Direct (NDE)	1.156***	(0.033)	0.988	(0.807)	1.122***	(0.031)	1.235***	(0.044)
Indirect (NIE)	1.011***	(0.002)	1.010***	(0.000)	1.001	(0.001)	1.004**	(0.002)
Total effect	1.169***	(0.033)	0.998	(0.969)	1.123***	(0.031)	1.239***	(0.044)
Mediate: Fear in the neighbourhood
Direct (NDE)	1.145***	(0.032)	0.971	(0.547)	1.109***	(0.031)	1.212***	(0.043)
Indirect (NIE)	1.026***	(0.003)	1.036***	(0.000)	1.016***	(0.002)	1.025***	(0.003)
Total effect	1.174***	(0.033)	1.006	(0.908)	1.126***	(0.031)	1.242***	(0.044)
No. of victims	6,769		2,051		7,588		4,128

NDE: natural direct effect; NIE: natural indirect effect; LTM: Last twelve months; ATM: After twelve months.

Discussion and conclusions

The literature has shown that the effects of violence may influence an individual's intention or decision to move (Dugan, 1999; Hipp et al., 2019; McNeeley & Stutzenberger, 2013; Xie & McDowall, 2008b; Yetter, 2018). In order to provide further empirical evidence, the present study examined if victimisation is associated with an increased desire to move in contemporary Brazil. We used a large and representative Brazilian survey of 61,787 individuals and employed a newly developed mediation analysis (Samoilenko & Lefebvre, 2021).

As proposed by McNeeley and Stutzenberger (2013), fear of crime was used to mediate the relationship between victimisation and the intention to move. In general, both personal crime (e.g., threat) and property crime (e.g., theft and robbery) were found to have a positive impact on fear of crime (aggression was the exception). These findings align with McNeeley and Stutzenberger (2013). They argue that personal victimisation, like aggression, could be a crime perpetrated by a known offender. Therefore, it does not affect victims’ perception of insecurity in their homes or neighbourhood.

Using the mediation analysis, we also tested four different models, as an individual's fear of crime affected by victimisation may influence the desire to move (McNeeley & Stutzenberger, 2013). Overall, it is worth noting that the indirect effects in our mediation analysis constitute a relatively small proportion of the Total Effect, which is consistent with similar studies (e.g., Samoilenko & Lefebvre, 2021). One possible reason for the relatively small indirect effect is that the survey might not accurately capture the fear level experienced by respondents. Another possible reason is that it might take some time for respondents to manifest significant levels of fear that will influence the desire to move, which may not be captured in cross-sectional survey data.

In our first model, we differentiate between the effects of property victimisation (robbery and theft) and person victimisation (aggression and threat). Consistent with the US studies focused on victimisation, we found that property victimisation plays a more important role in an individual's desire to move (Dugan, 1999). More specifically, our results indicate that theft victims (and not victims of robbery) have a greater desire to move, even though this is a less severe crime (McNeeley & Stutzenberger, 2013). These results differ from related studies investigating the effects of violent crimes (and not victimisation per se) and their attendant impact on the decision to move (e.g., Hipp, 2011).

However, through fear, victims of robbery may be affected. This appears to be the case when the violence occurs in or near the victim's residence, which is consistent with the findings of Xie and McDowall (2008b) and Dugan (1999). The reason is that the further away from home a particular crime occurs, an individual's desire to move is reduced.

Thirdly, we tested for indirect victimisation, or the knowledge of crime experienced by others, on an individual's desire to move. It was noted that our direct and indirect results for theft were broadly similar. However, although direct aggression does not significantly impact the desire to move, listening to or witnessing aggression seems relevant. As Xie and McDowall (2008b) argue, this may be because the victim of aggression usually knows the aggressor, which may be less likely in the case of indirect victimisation.

In our fourth model, we compared the effects of victims of property crime in the last 12 months to victims of property crime after 12 months. When the crime was theft, we found evidence that more recent victimisation was associated with a desire to move. Studies that consider more recent victimisation (Dugan, 1999; Xie & McDowall, 2008b) report similar results compared to those that consider longer periods (McNeeley & Stutzenberger, 2013). For robbery, direct and indirect effects present mixed results on the desire to move. Indirect effects of robbery (mediated by fear of crime) were higher for more recent victimisation (as observed for theft). However, the opposite result was found for the direct effects of robbery.

We presented two specifications for fear of crime: at home and at the neighbourhood level. Fear of crime in the neighbourhood, compared to the fear of crime at home, exerted a more substantial influence on the desire to move. One possible explanation is that individuals may have fewer mechanisms to protect themselves from violence at the neighbourhood level compared to violence at home, which alarm systems and security cameras may mitigate. Thus, fear of the wider environment could be a stronger driver of the desire to move.

Although this study contributes to previous research by confirming the effects of victimisation on the desire to move direct and indirectly through fear of crime, it has several limitations. First, although our models included a wide range of control variables, there are some important factors that we could not include due to data limitations (e.g., access to health care services). Although our results could be influenced by omitting relevant variables, we can take some comfort in the fact that we were able to control for a broad range of factors that may influence the desire to move.

Second, our results reflect moving intention. Since we used cross-sectional data, we analysed the desire to move, not actual moving behaviour. Therefore, our results may not be directly comparable to similar studies (e.g., Dugan, 1999; Xie & McDowall, 2008b). However, McNeeley and Stutzenberger (2013) also faced the same data limitation but contended that moving intention captures those individuals who cannot move due to limited financial resources.

A third possible limitation is related to the cross-sectional nature of our dataset. Our analysis may contain some individuals who reported not wanting to move from their residence, even when their response to past victimisation was positive. One possible reason is that they could have already moved from their violent neighbourhood, although we cannot identify it. However, this potential limitation could be verified and addressed using data collected from a longitudinal survey.

Another possible limitation is that our measure of fear might not be sufficiently sensitive to detect distinctions between crime types. This could possibly explain why fear at home and fear in the neighbourhood impact all crime types similarly (Table 2). On a related note, it could also take some time for these distinctions to manifest, which we might be able to identify in a longitudinal survey. In any case, this is an important consideration that is worthy of further reflection and investigation.

Despite these limitations, the present study showed a stronger desire to move among victims of violence. This, in turn, could lead to residential instability (i.e., people searching for a new place to live) (Tita et al., 2006), which may further expose individuals to victimisation (Miethe & Meier, 1996; Shaw & McKay, 1943; Xie & McDowall, 2008a). This could further exacerbate the psychological and financial costs of crime, victimisation, and residential mobility (Yetter, 2018).

Therefore, policymakers may be able to reduce the costs associated with crime by providing additional support services and community programmes to reduce victimisation and fear of crime. Moreover, when evaluating the associated costs of crime, policymakers may wish to explicitly consider the economic and social impact that crime has on moving (intended and actual). These costs may not always be included when conducting cost–benefit analyses on crime prevention and reduction programmes.

Footnotes

Acknowledgements

The authors want to thank Genevieve Lefebvre, Mariia Samoilenko, Trang Quynh Nguyen, Stacy Tiemeyer, and Karina Shreffler for their advice on the methodology.

Authors’ note

Luan V Bernardelli, The Special Academic Unit of Applied Social Sciences, Federal University of Goiás, Goiás - GO, Brazil.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Wander Plassa

Notes

References

Averdijk

(2011). Reciprocal effects of victimisation and routine activities. Journal of Quantitative Criminology, 27(2), 125–149. https://doi.org/10.1007/s10940-010-9106-6

Baydar

White

(1988). A method for analysing backward recurrence time data on residential mobility. Sociological Methodology, 18, 105–135. https://doi.org/10.2307/271046

Boggess

L. N.

Hipp

J. R.

(2010). Violent crime, residential instability and mobility: Does the relationship differ in minority neighbourhoods? Journal of Quantitative Criminology, 26(3), 351–370. https://doi.org/10.1007/s10940-010-9093-7

Cullen

J. B.

Levitt

S. D.

(1999). Crime, urban flight, and the consequences for cities. Review of Economics and Statistics, 81(2), 159–169. https://doi.org/10.1162/003465399558030

Dugan

(1999). The effect of criminal victimisation on a household’s moving decision. Criminology; An Interdisciplinary Journal, 37(4), 903–930. https://doi.org/10.1111/j.1745-9125.1999.tb00509.x

Engel

Ibáñez

A. M.

(2007). Displacement due to violence in Colombia: A household-level analysis. Economic Development and Cultural Change, 55(2), 335–365. https://doi.org/10.1086/508712

Fadaei-Tehrani

Green

T. M.

(2002). Crime and society. International Journal of Social Economics, 29(10), 781–795. https://doi.org/10.1108/03068290210444412

Ferraro

K. F.

(1997). Fear of crime interpreting victimisation risk. Social Forces, 76(1), 350–351. https://doi.org/10.2307/2580338

Fórum Brasileiro de Segurança Pública (2021). Anuário Brasileiro de Segurança Pública. Retrieved from https://forumseguranca.org.br/wp-content/uploads/2021/07/anuario-2021-completo-v4-bx.pdf

10.

Garofalo

(1981). The fear of crime: Causes and consequences. Journal of Criminal Law & Criminology, 72(2), 839–857. https://doi.org/10.2307/1143018

11.

Gaynor

S. M.

Schwartz

Lin

(2019). Mediation analysis for common binary outcomes. Statistics in Medicine, 38(4), 512–529. https://doi.org/10.1002/sim.7945

12.

Global Peace Index (2022). Global Peace Index 2022 report. Retrieved from https://www.visionofhumanity.org/wp-content/uploads/2022/06/GPI-2022-web.pdf

13.

Hipp

J. R.

(2011). Violent crime, mobility decisions, and neighbourhood racial/ethnic transition. Social Problems, 58(3), 410–432. https://doi.org/10.1525/sp.2011.58.3.410

14.

Hipp

J. R.

Tita

G. E.

Greenbaum

R. T.

(2009). Drive-bys and trade-ups: The impact of crime on residential mobility patterns in Los Angeles. Social Forces, 87(4), 1777–1812. https://doi.org/10.1353/sof.0.0184

15.

Hipp

J. R.

Williams

S. A.

Kim

Y. A.

Kim

J. H.

(2019). Fight or flight? Crime as a driving force in business failure and business mobility. Social Science Research, 82, 164–180. https://doi.org/10.1016/j.ssresearch.2019.04.010

16.

Lai

E. Y.

Shih

Huang

Y. T.

Wang

(2020). A mediation analysis for a nonrare dichotomous outcome with sequentially ordered multiple mediators. Statistics in Medicine, 39(10), 1415–1428. https://doi.org/10.1002/sim.8485

17.

Leibbrand

Gabriel

Hess

Crowder

(2021). When push comes to shove: Local violence and residential mobility among families. Journal of Marriage and Family, 83(5), 1249–1267. https://doi.org/10.1111/jomf.12769

18.

Liska

A. E.

Bellair

P. E.

(1995). Violent-crime rates and racial composition: Convergence over time. American Journal of Sociology, 101(3), 578–610. https://doi.org/10.1086/230754

19.

Mateyka

P. J.

(2015). Desire to move and residential mobility: 2010–2011. Review of Sociology, 28, 443–478. https://www.census.gov/content/dam/Census/library/publications/2015/demo/p70-140.pdf (accessed 26 April 2022)

20.

McNeeley

Stutzenberger

(2013). Victimisation, risk perception, and the desire to move. Victims & Offenders, 8(4), 446–464. https://doi.org/10.1080/15564886.2013.835295

21.

Miethe

T. D.

Meier

R. F.

(1996). Crime and its social context: Toward an integrated theory of offenders, victims, and situations. Social Forces, 74(4), 1449–1450. https://doi.org/10.2307/2580366

22.

Najdowski

C. J.

Ullman

S. E.

(2009). Prospective effects of sexual victimisation on PTSD and problem drinking. Addictive Behaviors, 34(11), 965–968. https://doi.org/10.1016/j.addbeh.2009.05.004

23.

Nguyen

T. Q.

Webb-Vargas

Koning

I. M.

Stuart

E. A.

(2016). Causal mediation analysis with a binary outcome and multiple continuous or ordinal mediators: Simulations and application to an alcohol intervention. Structural Equation Modeling: A Multidisciplinary Journal, 23(3), 368–383. https://doi.org/10.1080/10705511.2015.1062730

24.

Rø

A. D. B.

Simpson

M. R.

Rø

T. B.

Storrø

Johnsen

Videm

Øien

(2017). Reduced Th22 cell proportion and prevention of atopic dermatitis in infants following maternal probiotic supplementation. Clinical & Experimental Allergy, 47(8), 1014–1021. https://doi.org/10.1111/cea.12930

25.

Roth

Huffman

Brame

(2020). Too afraid to stay: Measuring the relationship between criminal victimisation in Central America and the intent to migrate. Crime & Delinquency, 68(4), 684–706. https://doi.org/10.1177/0011128720978737

26.

Rountree

P. W.

Land

K. C.

(1996). Burglary victimisation, perceptions of crime risk, and routine activities: A multilevel analysis across Seattle neighbourhoods and census tracts. Journal of Research in Crime and Delinquency, 33(2), 147–180. https://doi.org/10.1177/0022427896033002001

27.

Samoilenko

Blais

Lefebvre

(2018). Comparing logistic and log-binomial models for causal mediation analyses of binary mediators and rare binary outcomes: Evidence to support cross-checking of mediation results in practice. Observational Studies, 4(1), 193–216. https://doi.org/10.1353/obs.2018.0013

28.

Samoilenko

Lefebvre

(2021). Parametric-regression–based causal mediation analysis of binary outcomes and binary mediators: Moving beyond the rareness or commonness of the outcome. American Journal of Epidemiology, 190(9), 1846–1858. https://doi.org/10.1093/aje/kwab055

29.

SENASP (2012). National Secretariat for Public Security. Retrieved from https://www.crisp.ufmg.br/wp-content/uploads/2013/10/Relat%C3%B3rio-PNV-Senasp_final.pdf

30.

Shaw

C. R.

McKay

H. D.

(1943). Juvenile delinquency and urban areas. Harvard Law Review, 56(4), 681–681. https://doi.org/10.2307/1334446

31.

Sheikh

M. A.

Abelsen

Olsen

J. A.

(2017). Education and health and well-being: Direct and indirect effects with multiple mediators and interactions with multiple imputed data in stata. Journal of Epidemiology and Community Health, 71(11), 1037–1045. https://doi.org/10.1136/jech-2016-208671

32.

Shreffler

Tiemeyer

Giano

Gallus

Cox

Merten

(2020). Trauma and early adolescent perceptions about sex and parenthood: The mediating role of anger regulation. Youth & Society, 52(8), 1414–1435. https://doi.org/10.1177/0044118X18814051

33.

Skogan, W. G. (1992). Disorder and decline: Crime and the spiral of decay in American neighbourhoods. Univ of California Press.

34.

Skogan

Maxfield

(1982). Coping with crime: Individual and neighbourhood reactions. The Journal of Criminal Law and Criminology, 73(3), 1317–1319. https://doi.org/10.2307/1143199

35.

Tita

G. E.

Petras

T. L.

Greenbaum

R. T.

(2006). Crime and residential choice: A neighbourhood level analysis of the impact of crime on housing prices. Journal of Quantitative Criminology, 22(4), 299–317. https://doi.org/10.1007/s10940-006-9013-z

36.

Valeri

VanderWeele

T. J.

(2013). Mediation analysis allowing for exposure–mediator interactions and causal interpretation: Theoretical assumptions and implementation with SAS and SPSS macros. Psychological Methods, 18(2), 137. https://doi.org/10.1037/a0035596

37.

Warr

(1994). Public perceptions and reactions to violent offending and victimisation. In Reiss

A. J.

Jr. Roth

J. A.

(Eds.) & National Research Council, Understanding and preventing violence, vol. 4. Consequences and control (pp. 1–66). National Academy Press.

38.

Wood

Gibson

Ribeiro

Hamsho-Diaz

(2010). Crime victimisation in Latin America and intentions to migrate to the United States. The International Migration Review, 44(1), 3–24. https://doi.org/10.1111/j.1747-7379.2009.00796.x

39.

Xie

McDowall

(2008a). The effects of residential turnover on household victimisation. Criminology; An Interdisciplinary Journal, 46(3), 539–575. https://doi.org/10.1111/j.1745-9125.2008.00123.x

40.

Xie

McDowall

(2008b). Escaping crime: The effects of direct and indirect victimisation on moving. Criminology; An Interdisciplinary Journal, 46(4), 809–840. https://doi.org/10.1111/j.1745-9125.2008.00133.x

41.

Xie

McDowall

(2014). Impact of victimisation on residential mobility: Explaining racial and ethnic patterns using the national crime victimisation survey. Criminology; An Interdisciplinary Journal, 52(4), 553–587. https://doi.org/10.1111/1745-9125.12048

42.

Yetter

A. M.

(2018). Victimisation-precipitated residential mobility among women offenders. Crime & Delinquency, 64(13), 1718–1741. https://doi.org/10.1177/0011128717751663