Victimization-Precipitated Residential Mobility Among Women Offenders

Abstract

Victims of crime are likely to move residence following their victimization. However, the reasons for and the outcomes of victimization-precipitated moves remain unclear. The current study uses life event calendar data on jailed women to test two potential mechanisms: relationship dissolution and perceptions of neighborhood safety. In addition, this study seeks to understand how the safety of women’s residential contexts is affected by their past victimizations and residential mobility. Results show that intimate partner victimization is associated with increased odds of moving, and this relationship is partially mediated by relationship dissolution. Furthermore, moving and nonpartner victimization interact in their effects on neighborhood safety, such that moves following higher levels of victimization lead women into more dangerous neighborhoods.

Keywords

victimization residential mobility women offenders life event calendars

Introduction

Residential instability is a central concept in criminological explanations of neighborhood differences in crime rates (Kubrin, Stucky, & Krohn, 2009). When neighborhood residents are frequently moving out and replaced by new residents, the neighborhoods’ social cohesion and informal social control are undermined, which leads to increased crime rates (Bursik, 1988; Shaw & McKay, 1942). Furthermore, high crime rates act as a push factor causing many individuals to leave the neighborhood (Hipp, Tita, & Greenbaum, 2009). Consequently, many of those who remain in, or who move into, that high crime neighborhood are more likely to engage in crime and less likely to engage in informal social control (Stark, 1987). For these reasons, residential instability is cited as both a factor predictive of crime rates and as an outcome of high crime rates (South & Messner, 2000). In the aggregate, increases in neighborhood crime rates lead to out-migration and decreased populations in high crime areas, especially decreased white populations (Morenoff & Sampson, 1997). This relationship is largely explained by low perceptions of safety in the high crime areas, which causes many residents to move if they are able to (South & Messner, 2000). Implicit in studies of neighborhood-level trends are therefore individual-level experiences with crime and resulting residential mobility decisions.

Only a small body of research has studied the individual-level relationship between experiencing criminal victimization and subsequent residential mobility, but these studies find empirical support for the idea that victimization causes people to move (Dugan, 1999; Xie & McDowall, 2008b). Other studies have found that victimization is not only associated with a single move but also with more serious and lasting housing insecurity (Elklit & Shevlin, 2009; Pavao, Alvarez, Baumrind, Induni, & Kimerling, 2007), which indicates that postvictimization moves should perhaps be considered one of the many negative consequences of victimization. Whereas the assumption behind much of the current literature regarding crime and mobility is that moving following victimization is a protective strategy that is used to decrease the risk of repeat victimization, this positive outcome may not necessarily be the result. Empirical evidence indicates that mobility itself is associated with increased risk of victimization (Elklit & Shevlin, 2009; Waltermaurer, McNutt, & Mattingly, 2006; Xie & McDowall, 2008a) among other negative outcomes such as declines in health, mental health, and health care (Bures, 2003; Kushel, Gupta, Gee, & Haas, 2006) and increases in school dropout (Astone & McLanahan, 1994). Consequently, the current study seeks to understand the causes and consequences of victimization-precipitated moves.

Using 36-month life event calendar data from the Women’s Experience With Violence (WEV) study (Slocum, Simpson, & Smith 2005), the current study conducts a within-person, fixed-effects analysis of women offenders’ victimization, moves, and perceived neighborhood safety. My focus is threefold: First, I model the relationship between women’s violent victimization and the odds of a subsequent move. This relationship is also disaggregated to separately observe the associations of moving with intimate partner violence (IPV) and other violent victimization. Second, I include measures of relationship dissolution and perceived neighborhood safety to test whether these variables function as the mechanisms that explain the link between victimization and moving. Finally, I examine how women’s residential context is shaped by their past victimizations to understand whether moves following victimization are beneficial to victims’ neighborhood quality.

Background

Residential Mobility as a Consequence of Victimization

Experiencing victimization can cause many negative consequences, both short and long term, including compromised perceptions of personal agency and interpersonal support systems (Macmillan, 2001). Psychologically, victims may suffer from posttraumatic stress disorder (Bromet, Sonnega, & Kessler, 1998; Najdowski & Ullman, 2009), and may abuse alcohol and drugs (Kilpatrick, Acierno, Resnick, Saunders, & Best, 1997; Najdowski & Ullman, 2009). Violent victimization can, further, come at a high financial cost to victims who may have medical and mental health bills (Miller, Cohen, & Rossman, 1993). Moving residence is yet another consequence of victimization that may be quite costly for victims given the expenses involved in moving. In addition, the effect of aggregate neighborhood residential instability on crime rates means that patterns of postvictimization mobility may be quite costly for neighborhoods. Unlike posttraumatic stress disorder and alcohol and drug abuse, victimization-precipitated mobility has not been extensively researched, despite its potentially serious consequences for both victims and neighborhoods.

A small number of studies have begun to establish the empirical link between victimization and residential mobility (Burgess & Holmstrom, 1979; Dugan, 1999; Elklit & Shevlin, 2009; Pavao et al., 2007; Xie & McDowall, 2008b). In an early qualitative study of coping strategies used by rape survivors, Burgess and Holmstrom (1979) found that actively changing residence was a tool for victims to avoid reminders of their victimization. Many women achieved this change by moving in with friends or family, especially if they did not have many financial resources. The moves were described positively by the victims, and the authors categorize them as adaptive, rather than maladaptive, coping strategies.

Later studies used representative data from the National Crime Survey (NCS) to test whether households are likely to move following victimization. Dugan (1999) found that victimization within 1 mile from home was significantly associated with moving. Specifically, Dugan (1999) found that nearby property, but not violent, victimizations were associated with a subsequent move. Xie and McDowall (2008b) expanded upon Dugan’s (1999) work, also using the NCS. They not only studied the association between household victimization and moving but also included neighbors’ victimization as a measure of the impact of indirect victimization on moving. This study found that, controlling for neighborhood characteristics, households’ direct experiences with property and violent crime increase the odds of a subsequent move. Furthermore, Xie and McDowall (2008b) found that households were more likely to move when their adjacent neighbors experienced property victimization but not violent victimization. Together, these studies identify victimization as an important predictor of residential mobility.

Research has also examined how IPV victimization and sexual victimization contribute to individuals’ housing insecurity. By studying housing insecurity, this study did not only examine moves but also used a broader measure that included being more than 30 days late on rent or mortgage payments, not having housing at some point in the year, or making more than one move in a year. Using data from a random-digit-dial survey of Californian women, Pavao and colleagues (2007) found that IPV in the past year was associated with a fourfold increase in the odds of housing insecurity in the same time period. Using a similar measure of housing insecurity, a study of Danish women found that those who had visited a rape crisis center were more likely than a matched sample of women to move municipalities 2 or more times in the following 3 years (Elklit & Shevlin, 2009). Together, this small body of research provides evidence of a relationship between victimization and housing but describes two very different pictures of the reason for that relationship.

Mechanisms of the Victimization–Mobility Relationship

Overall, the quantitative research in this area has thus far demonstrated that residential mobility is one noteworthy outcome of victimization experiences at the individual and household levels, but this work has been limited in its ability to assess the mechanisms that explain this relationship. The data sources that provide information about victimization and mobility often do not measure fear of crime or other variables that may mediate the association between victimization and subsequent moves. Moreover, the different operationalizations of residential mobility in these studies are conceptually distinct, and may imply different mechanisms leading to moves and different consequences of moves. Specifically, the nationally representative studies using the NCS data measure residential mobility by a single move, and cannot assess whether a household experiences chronic housing insecurity characterized by repeated moves or difficulty paying rent. The qualitative experience implied by measuring one move is distinct from that implied by measuring repeated moves or difficulty maintaining housing. To draw conclusions about the meaning of victimization-precipitated moves, the mechanisms need to be known.

Generally, the assumption behind studies examining a single move is that victimization results in a decreased perception of neighborhood safety, and the solution is a subsequent move to a new neighborhood. These moves would therefore be indicative of the adaptive coping strategies identified by Burgess and Holmstrom (1979). This perspective fits with aggregate studies’ depiction of White residents’ moves away from urban high crime areas (Morenoff & Sampson, 1997). One study (McNeeley & Stutzenberger, 2013) has tested this proposed mechanism using data from the Seattle Neighborhoods and Crime Survey and found that property victimization indirectly increased the desire to move by increasing respondents’ perceptions of risk. This study was not able to measure actual moving behavior, and previous studies have found that the reported desire to move is not necessarily related to actual moving behavior (Newman & Duncan, 1979). As such, no study has yet examined perceived neighborhood safety as the mechanism connecting victimization to actual moving behavior.

Studies that examine the broader concept of housing insecurity, however, argue that victimization is related to a number of negative outcomes that may also lead to the inability to maintain a stable residence. For example, in the case of sexual victimization, research has found lower relationship quality and decreased social/relational adjustment as a result of victimization (Zweig, Crockett, Sayer, & Vicary, 1999). These changes could lead victims to end relationships with partners, even if they were not victimized by their partner, which may result in moving. Alternatively, ending a relationship and leaving an abusive partner is certainly one likely mechanism whereby victims of IPV come to experience residential mobility. Studies of women who have left violent relationships find that housing problems persist after women move for a number of reasons, including persistent harassment from the former partner and financial problems that make it difficult to pay rent (Baker, Cook, & Norris, 2003). Financial problems are a common outcome of IPV, which can limit women’s access to employment and education (Pavao et al., 2007). These descriptions of postvictimization moves are qualitatively different than those assumed to be driven purely by relocating to a new, safer neighborhood.

Consequences of Mobility

In considering the possible stories behind victimization-precipitated moves, it is important to understand what consequences mobility may have, and how it is restricted. Moving may be a positive as well as a negative outcome. For many people, moving may well result in “Escaping Crime,” as Xie and McDowall (2008b) titled their study. The finding that victimization leads to an increased perception of risk (McNeeley & Stutzenberger, 2013) supports the idea that the intended consequence of victimization-precipitated moving is finding a safer neighborhood in which to live. Ideally all victims who seek such an outcome would achieve it, but no study to date has empirically studied the kinds of moves made by victims. The literature cannot currently speak to the destinations of victims who move.

Studies that use housing insecurity as their outcome assume a different possible story. They make the assumption that the mobility-related consequences of victimization will not be beneficial (Elklit & Shevlin, 2009; Pavao et al., 2007). Although these studies do not measure the characteristics of victims’ postmove neighborhoods, this perspective views victimization as financially destabilizing in ways that prevent the maintenance of a steady residence, which also likely prevents attaining improved neighborhood quality. Escaping crime seems much less likely under these circumstances. Indeed, research has found that residential mobility increases the risk of subsequent victimization (Elklit & Shevlin, 2009; Waltermaurer et al., 2006; Xie & McDowall, 2008a). Moreover, victims of IPV who struggle to maintain reliable housing after leaving their abuser may be more likely to return to the violent relationship, furthering their risk of victimization (Pavao et al., 2007). This portrayal of victimization-precipitated residential mobility indicates that these moves are a point of vulnerability. The extent to which a victim is negatively affected by a move may be further stratified.

Studies of mobility find that the likelihood of moving is not evenly distributed across the population. A number of sociodemographic factors influence the ability to move and where moves occur. There is some contradicting evidence in this area, with some reports that Whites are less mobile than non-Whites (Schachter, 2004) and some reports that Blacks are less mobile than non-Blacks (South & Deane, 1993). The aggregate studies of out-migration from high crime areas find that there is variability by race. Morenoff and Sampson (1997) found that White populations decreased and Black populations increased when neighborhood homicide and proximity to homicide increased, which indicates that Blacks have more difficulty escaping crime than Whites. Furthermore, studies have found that Blacks move into disadvantaged census tracts more often and are less likely to move out of them (South & Crowder, 1997). In addition, Blacks are less likely to benefit from capital that otherwise helps other racial and ethnic groups to move into advantaged areas (Logan & Alba, 1993). When this evidence is seen alongside findings that indicate that individuals with low and high incomes are more likely to move than those with middle income (Xie & McDowall, 2008b), there is a two-part story of residential mobility. Specifically, it seems that more advantaged people (non-Black, higher socioeconomic status [SES]) are less restricted in their mobility. They can more easily move out of disadvantaged neighborhoods. Blacks and people with lower SES, however, are making moves that appear more characteristic of housing insecurity than just mobility, and are likely restricted to disadvantaged areas.

The current study uses a sample of jailed women, many of whom are minorities, and have low SES and educational attainment. Through 36-month retrospective life event calendar data, the study is able to trace these women’s victimization and residential histories. This study contributes to the literature by testing the following hypotheses:

Hypothesis 1: Victimization will increase the odds of a subsequent move.

Hypothesis 2: IPV and nonpartner violence will both increase the odds of a subsequent move.

Hypothesis 3: Relationship breakup will mediate the effect of IPV on moving.

Hypothesis 4: Perceived neighborhood safety will mediate the effect of nonpartner victimization on moving.

Hypothesis 5: Finally, given the demographic characteristics of this sample, victimization and mobility are expected to interact to decrease neighborhood quality.

Data and Method

Data

The data for this article come from the WEV study, which conducted in-person interviews with 825 women who were incarcerated in Baltimore, Minneapolis, and Toronto (Slocum et al., 2005). The goal of the WEV study was to understand women’s experiences of violence, as both victims and offenders, in the specific circumstances of their lives. Female graduate student interviewers used computerized life event calendars to gather information about the women’s lives over the 36 months prior to becoming incarcerated, including their economic, social, family, and living situations. Also collected were time-invariant data on the respondent’s demographics, family background, and offense history. The current study focuses on the 36-month life event calendars to understand how women offenders’ victimization experiences affected their residential mobility over this time period.

Measures

Dependent variables

I use two different dependent variables in this study: First, I model whether the respondent changed residence. A respondent is coded as having moved if the residential address she reported for a given month is different from the previous month’s address.¹ Addresses were reported as the nearest cross-street. This is a dichotomous measure where a month with a move is coded “1” and a nonmove “0.” Movement in and out of in-patient treatment centers and incarceration was not coded as a move.

The second dependent variable is a measure of the perceived safety of respondents’ residential neighborhoods. This variable is a scale composed of three different variables (α = .78): First, respondents were asked, “In each month, in your neighborhood, how safe was it for women and children to walk around at night?” They could answer “very safe,” “fairly safe,” “fairly unsafe,” or “very unsafe.” Second, respondents were asked about the frequency of gunshots in their neighborhood. They were specifically asked, “In each month, how often did you ever hear gunshots in this neighborhood?” They could answer “never,” “once or twice a month,” “at least once a week,” or “almost every day.” Finally, respondents were asked, “In each month, how often did you ever witness violent attacks on persons outside your home in this neighborhood?” The possible responses for this variable were the same as those for the gunshots variable. The standardized scale of these variables is the dependent variable for this study measuring perceived neighborhood safety. Higher values on this scale represent more safety.

Independent variables

I operationalize the key independent variable of violent victimization in three different ways: The first is an overall measure of victimization. This is a variety score of victimization in each month, constructed by taking the sum of five dichotomous variables for whether the respondent was (1) attacked by her partner, (2) forced to have sex with her partner, (3) robbed by a nonpartner, (4) forced to have sex with a nonpartner, or (5) attacked by a nonpartner. The second and third independent variables of interest separate out the above measures into separate variety score variables of IPV victimization (1 and 2) and nonpartner victimization (3-5). Each of these variety scores was lagged across the past 4 months, and the average was taken across those four lags, making the final independent variables an average variety score over the 4 months prior to the month of the move. In other words, in each month the respondents reported experiencing between 0 and 5 types of violent victimization, two of which could be IPV and three of which could be nonpartner violence. These variables are represented by an average of the total number of these victimizations that the respondent experienced over the past 4 months. Variety scores were chosen because they capture more variation than a dichotomous measure, and they are highly reliable and valid (Sweeten, 2012). The 4-month lag was chosen because residential moves can rarely happen suddenly. Time is required to find a new home and to end a lease or sell a previous home. Sensitivity analyses, discussed below, were conducted using 6-month and 2-month lags and dichotomous victimization measures.

Control variables

I draw control variables primarily from research that explains the influences of individual and family characteristics on residential mobility. Like the victimization variables, the control variables are an average of the 4 months prior to the residential move. Prior research finds that unemployment can lead to a move (Herzog & Schlottmann, 1984; Xie & McDowall, 2008b), whereas income has a less clear relationship with moving. Mobility may be more likely among individuals with both high (Dugan, 1999) and low (Schachter, 2004) incomes than among those with middle incomes (Xie & McDowall, 2008b). However, the quality of the moves made by high- and low-income individuals will likely be quite different based on their resources. For my study, I include a dichotomous measure of whether the respondent was employed (coded “1”) prior to the move. There is only a valid income measure for respondents who were employed, making income highly collinear with the employment measure. Instead, as a proxy for low income and financial hardship, I include a dichotomous measure of whether the respondents reported any “legal income from Food Stamps, TANF, Social Security, etc.” (coded “1”), which I call other income.

Family development is also related to mobility. More specifically, there are lifecycle stages in families that influence residential moves. Age affects residential stability because young adults are more likely to experience transitions that bring about moves, such as marriage (Schachter, 2004). Being married (coded “1”), however, increases residential stability (Xie & McDowall, 2008b). At the same time, I control for whether the respondent’s relationship with her intimate partner ended in the month leading up to the move (this is also tested as a mediating variable, relationship dissolution coded “1”). Both married and relationship dissolution are dichotomous indicators. Households with children (Long, 1972) and larger households (Xie & McDowall, 2008b) are less likely to move. I include measures of the number of biological children the respondent has as well as whether the respondent is the primary caregiver (coded “1”) for any people other than her children. A final household factor that may influence moving behavior is residential duration, with longer tenure decreasing moves (Speare, 1974). Although I cannot measure true residential duration because the housing information is limited to the 36-month period, I control for the proportion of the previous 4 months where the respondent had a prior move. I also control for the respondent’s perceived neighborhood safety (the scale described above) in the month immediately prior to the move.

Finally, because the respondents are all offenders, I control for a variety score of offending that is constructed by taking the mean number of different offenses committed in each month. Offending behaviors are dichotomous indicators of burglary, theft, autotheft, prostitution, fraud, forgery, drug dealing, robbery, assault, rape, partner rape, and partner assault. This variable, like the others, is an average of the previous 4 months.

Analytic strategy

Two types of models are included in this article: First, I use logistic regression to model whether the respondent moved. Logistic regression is appropriate for these models because the dependent variable is dichotomous. The results reported in these tables represent the change in the logged odds of moving, given a 1-unit increase in the independent variable. Second, I use ordinary least squares (OLS) regression to model the perceived safety of the respondents’ neighborhoods because this scale is roughly continuous, ranging from −2.05 to .94. The results reported in these tables represent the increase in the Perceived Safety scale associated with a 1-unit increase in the independent variable.

All analyses are conducted using fixed-effects estimation which uses each respondent as her own control. This estimation strategy allows for conclusions that approximate causality because unmeasured time-stable characteristics of the respondents cannot bias the estimates. Due to this estimation strategy, only time-varying measures need to be included in the models because time-stable characteristics are controlled via the within-person design. Accordingly, some variables that prior research has found to have an important impact on mobility, such as education and race (Logan & Alba, 1993; Schachter, 2004; South & Crowder, 1997; South & Deane, 1993), are not included in the models. The 36-month life event calendar data are suited to the use of within-person analysis because they capture a large number of “waves” within which the variables of interest may vary over the respondents’ lives. The number of waves is reduced in my analysis because the independent variables are lagged over 4 months, so the first 4 months cannot be included in the analysis due to missing data (i.e., the months prior to the 36-month study period). The models are estimated using months 5 to 36. Descriptive information for the included variables is presented in Table 1.

Table 1.

Descriptive Statistics of Months.

	Overall				Moved				No moves
	M	SD	Minimum	Maximum	M	SD	Minimum	Maximum	M	SD	Minimum	Maximum
Moved	0.06	0.24	0	1	0.09	0.29	0	1	0	0	0	0
Perceived neighborhood safety	0.00	0.83	−2.05	0.94	0.01	0.84	−2.05	0.94	−0.01	0.80	−2.05	0.89
Total victimization	0.10	0.32	0	5	0.12	0.35	0	5	0.06	0.24	0	2.25
IPV victimization	0.07	0.27	0	2	0.09	0.30	0	2	0.04	0.21	0	2
Nonpartner victimization	0.02	0.15	0	3	0.03	0.17	0	3	0.02	0.10	0	1.25
Relationship dissolution	0.02	0.15	0	1	0.03	0.16	0	1	0.01	0.12	0	1
Offending	0.06	0.08	0	0.69	0.06	0.08	0	0.44	0.06	0.08	0	0.69
Employed	0.35	0.46	0	1	0.36	0.46	0	1	0.33	0.45	0	1
Other income	0.34	0.47	0	1	0.34	0.46	0	1	0.35	0.47	0	1
Caregiver	0.05	0.21	0	1	0.05	0.20	0	1	0.05	0.21	0	1
Married	0.16	0.37	0	1	0.17	0.37	0	1	0.15	0.35	0	1
Children	2.02	1.71	0	9	1.98	1.71	0	9	2.12	1.70	0	9
Prior moves	0.06	0.14	0	1	0.09	0.16	0	1	0	0	0	0

Note. IPV = intimate partner violence.

Not all of the 825 women surveyed moved during the 36-month calendar period. Fixed-effects regression cannot be conducted for respondents with no variation on the dependent variable, which removed 324 women from the sample in the logistic regression models predicting moving. Missing data are casewise deleted, which removes a further 10 women from these models. All respondents had some change in their reports of perceived neighborhood safety; hence, none were dropped from the OLS regressions for lack of variation. An additional three women, for a total of 13, were casewise deleted in the OLS models.

Results

Moving

Does victimization lead to moving? Tables 2 and 3 present the results of fixed-effects logistic regressions predicting the log odds of moving. The key independent variables are the variety scores of all victimization (Table 2) and IPV and nonpartner victimization (Table 3). Each of these tables includes four models: The first model includes the victimization measure along with the time-varying control variables. The two potential mediating variables are excluded. The next models (in columns 2 and 3) add in lagged perceived neighborhood safety and relationship dissolution separately as tests of mediation of the victimization to moving effect. Finally, column 4 includes the full model with both mediators included.

Table 2.

Fixed-Effects Logistic Regression Predicting Moving (vs. Not Moving), N = 13,202 Months, 491 Respondents.

	Coefficient	Coefficient	Coefficient	Coefficient
	(SE)	(SE)	(SE)	(SE)
Victimization	0.288^†	0.297*	0.265^†	0.275^†
	(0.150)	(0.150)	(0.150)	(0.150)
Neighborhood safety		0.124*		0.130*
		(0.060)		(0.060)
Relationship dissolution			0.881***	0.886***
			(0.140)	(0.140)
Offending	−3.302***	−3.025***	−3.320***	−3.030***
	(0.774)	(0.786)	(0.774)	(0.786)
Employed	−0.059	−0.083	−0.060	−0.084
	(0.115)	(0.116)	(0.115)	(0.116)
Other income	0.109	0.110	0.115	0.116
	(0.143)	(0.143)	(0.143)	(0.143)
Primary caregiver	−0.130	−0.103	0.143	−0.101
	(0.285)	(0.285)	(0.285)	(0.285)
Married	0.258	0.262	0.242	0.248
	(0.292)	(0.292)	(0.294)	(0.293)
Children	0.104	0.115	0.082	0.094
	(0.168)	(0.168)	(0.169)	(0.169)
Prior moves	0.469**	0.477**	0.477**	0.486**
	(0.157)	(0.157)	(0.157)	(0.157)

†

p < .1. *p < .05. **p < .01. ***p < .001.

Table 3.

Fixed-Effects Logistic Regression Predicting Moving (vs. Not Moving), N = 13,202 Months, 491 Respondents.

	Coefficient	Coefficient	Coefficient	Coefficient
	(SE)	(SE)	(SE)	(SE)
IPV victimization	0.475**	0.467**	0.442*	0.433*
	(0.179)	(0.179)	(0.179)	(0.179)
Nonpartner victimization	−0.172	−0.128	−0.166	−0.119
	(0.286)	(0.287)	(0.286)	(0.287)
Neighborhood safety		0.115^†		0.122*
		(0.060)		(0.060)
Relationship dissolution			0.876***	0.881***
			(0.140)	(0.140)
Offending	−3.196***	−2.949***	−3.220***	−2.959***
	(0.779)	(0.790)	(0.778)	(0.789)
Employed	−0.080	−0.100	−0.079	−0.100
	(0.116)	(0.116)	(0.116)	(0.116)
Other income	0.100	0.102	0.107	0.109
	(0.143)	(0.143)	(0.143)	(0.144)
Primary caregiver	−0.134	−0.108	−0.133	−0.106
	(0.286)	(0.286)	(0.286)	(0.285)
Married	0.257	0.262	0.243	0.248
	(0.292)	(0.292)	(0.294)	(0.293)
Children	0.110	0.120	0.088	0.099
	(0.168)	(0.168)	(0.169)	(0.169)
Prior moves	0.466**	0.474**	0.474**	0.483**
	(0.157)	(0.157)	(0.157)	(0.157)

Note. IPV = intimate partner violence.

†

p < .1. *p < .05. **p < .01. ***p < .001.

Table 2 shows moderate evidence of a relationship between victimization in the previous 4 months and the log odds that the respondent moves. In only one model is the coefficient for victimization significant at the p < .05 level. This finding is notably weaker than that found in previous research on victimization-precipitated mobility. It may be the case that the characteristics of the sample affect the overall relationship between victimization and moving. In particular, offending is significantly associated with a large decrease in the odds of moving, with an odds ratio of 0.037. The evidence of some positive association of victimization is noteworthy, given that such a strong effect of offending prevents moving. Finally, prior moves were also associated with an increase in the odds that the respondent would move. A higher number of moves in the past 4 months increased the odds of moving by a factor of 1.6, indicating that a history of unstable housing is likely to continue. When the hypothesized mediators are added to the model, both are significantly associated with moving. An increase in the neighborhood safety increases the odds of moving by a factor of 1.13, which indicates greater mobility in safer neighborhoods. In addition, relationship dissolution increases the odds of the respondent moving by 2.41 times.

In Table 3, the overall victimization measure is disaggregated into IPV and nonpartner victimization. Of these measures, only the IPV victimization measure is significant and indicates that an increase in experiences of IPV victimization increases the odds of a move by 1.61 times. Entering perceived neighborhood safety does not mediate the effect of IPV victimization on moving. In column 3, when relationship dissolution is entered into the model, the effect size of IPV victimization is slightly attenuated (from an odds ratio of 1.61-1.56), and the significance reduces from p < .01 to p < .05. This change indicates that a relationship ending explains some, but not all, of the association between IPV victimization and moving. The effect size of breakup is comparable in these models with those in Table 2. The pattern of the control variables in this table is substantively the same as in Table 2.

Perceived Neighborhood Safety

Do moves improve victims’ perceived neighborhood safety? The second set of models is presented in Tables 4 and 5. These tables present fixed-effects OLS regressions of perceived neighborhood safety. These models include the same variables from the full model in Tables 1 and 2 with the addition of the measure of whether a move occurred in that month (the dependent variable in Tables 2 and 3) as an additional predictor variable. The dependent variable of safety is measured in the same month as the move. In other words, the dependent variable is the perceived safety level of the respondent’s new neighborhood, if she has moved. In the second column of each of these tables, victimization is interacted with moving.

Table 4.

Fixed-Effects OLS Regression Predicting Perceived Neighborhood Safety, N = 20,219 Months, 812 Respondents.

	Coefficient	Coefficient
	(SE)	(SE)
Moved	−0.028***	−0.025***
	(0.006)	(0.006)
Victimization	−0.006	−0.003
	(0.008)	(0.008)
Moved × Victimization		−0.030^†
		(0.018)
Lag neighborhood safety	0.867***	0.867***
	(0.004)	(0.004)
Relationship dissolution	−0.007	−0.007
	(0.010)	(0.010)
Offending	0.036	0.035
	(0.044)	(0.044)
Employed	0.004	0.004
	(0.006)	(0.006)
Other income	0.003	0.003
	(0.008)	(0.008)
Primary caregiver	0.030^†	0.030^†
	(0.016)	(0.016)
Married	0.007	0.007
	(0.016)	(0.016)
Children	−0.008	−0.008
	(0.009)	(0.009)
Prior moves	−0.030*	−0.031**
	(0.012)	(0.012)
Constant	0.012	0.013
	(0.019)	(0.019)

Note. OLS = ordinary least squares.

†

p < .1. *p < .05. **p < .01. ***p < .001.

Table 5.

Fixed-Effects OLS Regression Predicting Perceived Neighborhood Safety, N = 20,219 Months, 812 Respondents.

	Coefficient	Coefficient
	(SE)	(SE)
Moved	−0.029***	−0.023***
	(0.006)	(0.006)
IPV victimization	0.009	0.007
	(0.010)	(0.010)
Nonpartner victimization	−0.050**	−0.026
	(0.016)	(0.017)
Moved × IPV Victimization		0.015
		(0.019)
Moved × Nonpartner Victimization		−0.321***
		(0.046)
Lag neighborhood safety	0.866***	0.867***
	(0.004)	(0.004)
Relationship dissolution	−0.007	−0.007
	(0.010)	(0.010)
Offending	0.043	0.040
	(0.044)	(0.044)
Employed	0.003	0.002
	(0.006)	(0.006)
Other income	0.002	0.002
	(0.008)	(0.008)
Primary caregiver	0.028^†	0.029^†
	(0.016)	(0.016)
Married	0.007	0.005
	(0.016)	(0.016)
Children	−0.009	−0.008
	(0.009)	(0.009)
Prior moves	−0.031**	−0.031**
	(0.012)	(0.012)
Constant	0.014	0.012
	(0.019)	(0.019)

Note. IPV = intimate partner violence; OLS = ordinary least squares.

†

p < .1. *p < .05. **p < .01. ***p < .001.

In Table 4, it is shown that victimization is not a significant predictor of the respondents’ levels of perceived neighborhood safety. Moving, however, is a significant predictor. The coefficient indicates that respondents who have moved, compared with those who have not, are .03 points lower on the Perceived Neighborhood Safety scale, indicating that moving leads respondents to neighborhoods that they perceive to be more dangerous than their origin neighborhoods. Also significant in this model is the lagged measure of perceived neighborhood safety, which has a large effect size given its close relationship with the dependent variable. For each unit increase in the past 4 months’ perceived neighborhood safety, the current month’s neighborhood is .87 units higher (i.e., safer). Finally, prior moves are significantly associated with a decrease in perceived neighborhood safety, indicating that women with a history of housing insecurity live in neighborhoods they find to be more dangerous. The interaction between victimization and moving is only marginally significant in this model.

In Table 5, the victimization measure is disaggregated into IPV and nonpartner violence. In the first model, a 1-unit increase in nonpartner victimization is significantly associated with a decrease in perceived neighborhood safety of .05 units. IPV victimization is not significantly associated with perceived neighborhood safety. Again, moving, lagged perceived neighborhood safety, and prior moves show significant associations with neighborhood safety comparable with those shown in Table 4. In the second model shown in Table 5, moving and nonpartner victimization negatively and significantly interact in their effects on perceived neighborhood safety.

This interaction is best understood through the graph shown in Figure 1. The x-axis represents increasing levels of victimization (measured as the average of the variety score over 4 months), and the y-axis is perceived neighborhood safety with higher values representing more safety. The two lines represent the difference in the effect of victimization on perceived safety for those who do and do not move. In this figure, for respondents who move, moving in association with experiencing a greater number of types of victimization leads to progressively more dangerous neighborhoods. In contrast, for those who do not move, increasing victimization has only a slight decrease in perceived neighborhood safety. This difference indicates that it is unlikely that experiencing victimization changes only perceptions of neighborhood safety, but that respondents do indeed move into less safe neighborhoods.

Figure 1.

Interaction of moving and victimization on perceived neighborhood safety.

Sensitivity Analyses

Two aspects of the models were adjusted to test the sensitivity of the results: First, I manipulated the size of the lag for the victimization and control variables. When the lag is extended to include 6 months of information, there is a significant negative effect of employment on moving. This indicates that it may be only relatively long periods of employment that have a stabilizing influence on residence. In addition, in the 6-month lag models, the effect of prior moves on moving changes direction. Whereas in the 4-month models prior moves increase the odds of another move, in the 6-month models prior moves decrease the odds. It may be the case that spells of residential instability are of limited lengths, captured by the 6-month lag. The 2-month lag models are largely consistent with the 4-month lag. However, in the 2-month model, the effect of neighborhood safety on moving drops from significance. In addition, the 2-month lag finds a significant positive interaction between IPV victimization and moving. In other words, IPV victimization-precipitated moves lead respondents into neighborhoods they perceive as more safe but only for moves that follow relatively shortly after the victimization experience. This effect may be explained by the circumstances of IPV victimization-precipitated moves. These moves may be into shelters or with family, both of which may be located in relatively more advantaged neighborhoods.

Second, I tested whether a dichotomous measure of victimization would yield different results from the variety score. Overall, substantive conclusions remain the same. The marginally significant effect of total victimization on moving drops out of significance, but the strong effect of IPV victimization remains the same. An additional interaction appears between moving and total victimization, indicating that victimization-precipitated moving leads to a decrease in neighborhood safety. Given that this interaction does not appear in both the variety score and dichotomous models, I do not draw any conclusions from this finding, but it is substantively consistent with the finding regarding nonpartner victimization in the variety score models, which remains in the dichotomous models.

Discussion

Prior research has documented residential mobility and housing insecurity following victimization (Dugan, 1999; Elklit & Shevlin, 2009; Pavao et al., 2007; Xie & McDowall, 2008b). Considering the psychological and financial cost of victimization and moving for victims, and the positive effect of neighborhood levels of residential instability on crime rates, the link between victimization and mobility is an important relationship to research more fully. To do so, the current study examines 36-month retrospective life event calendar data from the WEV study that was collected through interviews with jailed women, a population particularly vulnerable to housing instability and victimization.

Findings do not fully confirm previous research. Logistic regressions indicated that IPV victimization has a substantial influence on moving, but effects for overall victimization fell just outside of significance and nonpartner victimization was not associated with moving. There are a number of potential reasons for these findings. While prior studies have found violent victimization to lead to moving (Elklit & Shevlin, 2009; Pavao et al., 2007; Xie & McDowall, 2008b), Dugan (1999) only found an effect on moving for property victimization. Dugan (1999) proposed that violent victimization may be addressed through other means, such as changing routine activities, more easily than can property crime. My finding that IPV victimization but not nonpartner victimization is associated with moving fits with this expectation. For women victimized by intimate partners, the only change likely to address the violence may be ending the relationship and moving away from the abuser. In accordance with this idea, relationship dissolution partly mediated the association between partner violence and moving.

Another potential explanation for the differences between prior research and the current study is the sample. The WEV study surveyed jailed women. Accordingly, the respondents have a history of criminal offending, which has been documented to also affect residential mobility. Studies of the formerly incarcerated have found evidence of high rates of housing insecurity (Herbert, Morenoff, & Harding, 2015) primarily due to the terms and sanctions of parole supervision (Harding, Morenoff, & Herbert, 2013). It is also common for former offenders to rely on female relatives for housing when they are not able to secure it for themselves (Western, Braga, Davis, & Sirois, 2015), which can prove unstable. Indeed, the findings of the current study indicate a strong effect of offending on residential mobility; however, it runs counter to expectations from the prior literature. Results show that offending has a strong negative effect on moving, which indicates that when women are engaged in more criminal activities, they are less likely to change residence. Despite this strong countervailing offending effect, IPV victimization still prompts women to move. Further research is needed into the interplay between these two experiences in women’s lives and the role of the victim–offender overlap in shaping mobility.

The current study also examines the consequences of victimization-precipitated moves. The association between victimization and moving has been framed in two primary ways by prior research. First, some studies have discussed victimization-precipitated moves as “escaping crime” (Xie & McDowall, 2008b). This perspective frames these moves as a sort of adaptive coping mechanism to avoid repeat victimization. However, research on housing insecurity has argued that victimization leads to repeated moving and other signs of housing insecurity such as difficulty paying rent. This perspective frames these moves as problematic consequences of victimization. To test the assumptions behind these two perspectives, the current study predicts women’s perceived neighborhood safety. The results indicate that nonpartner victimization and moving interact, such that moves following higher levels of victimization lead to much more dangerous new neighborhoods.

It is also important to note about the interaction graphs that the level of perceived neighborhood safety is stable across levels of IPV and nonpartner victimization when respondents do not move. This is inconsistent with previous research, which found that victimization decreases individuals’ perceptions of their neighborhood’s safety (McNeeley & Stutzenberger, 2013). Because the measure of moving used in this study does not necessarily imply a change of neighborhood (i.e., it includes within- and between-neighborhood mobility), it could be unclear whether the decrease in perceived neighborhood quality resulted from the move or the victimization. The flat line for those who do not move provides evidence that this is not the case. Rather, in this sample of women, residential moves following victimization result in relocation to neighborhoods perceived as more dangerous, while victimization does not change perceived neighborhood safety when there is no move.

Like any research, this study has limitations: First, the study is limited by the number of measures in the dataset. The study does not have a measure of property victimization, which limits study conclusions to the effects of violent victimization only. Future studies should replicate these analyses with information about both property and violent victimization. The study also does not have a measure of homeownership, which is an important predictor of residential mobility because moving is more challenging for homeowners (Speare, 1974; Xie & McDowall, 2008a). However, the study is able to account for a substantial number of other factors related to moving decisions.

Second, one may question the accuracy of retrospective data based on the potential flaws in respondents’ memories. However, the method of collecting life event calendar data is designed to capitalize on what is known about how memory works. For example, respondents first plot major life events, which can serve as useful reference points. In their review of studies on the quality of life event calendar data, Roberts and Horney (2010) found that the data collected are often found to be of higher quality than data collected from standard survey methods, with very few studies finding it to be of lower quality than standard survey data. However, some characteristics of this sample (such as experiences with violence) have been found to decrease recall.

A final limitation is that the sample is not generalizable to the population as a whole. However, the population from which this sample is drawn is highly vulnerable to victimization and mobility. To measure enough violent victimization to analyze, a nationally representative sample would have to be very large, which would prevent the collection of the detailed life event calendar data that are included in the WEV study. Nevertheless, the population to which this sample can generalize is one that is likely to be the target of policies for rehabilitation, making the findings of this research and their implications highly policy relevant regardless of generalizability.

Conclusion

Overall, the findings of this study indicate that IPV victimization is significantly associated with increased odds of subsequent moving. This relationship is partly mediated by the dissolution of that partnership. Second, this study finds that moving and nonpartner victimization interact, which indicates that moving leads to progressively more dangerous residential neighborhoods following increased experiences of victimization. Contrary to the assumptions of some studies, this finding implies that victimization-precipitated moving is not necessarily a successful strategy to use in response to increased perceptions of neighborhood risk. Instead, this finding fits more closely with the assumptions of studies of housing insecurity that view moving after victimization as one of many negative consequences of victimization. Given the demographic characteristics of the respondents included in this sample, the negative consequences of postvictimization moves may be associated with the respondents’ other disadvantages. Many of these women are minorities with low SES, as well as offenders. These characteristics have been linked to restrictions on where individuals move. If indeed the moves made by more advantaged populations improve their neighborhood safety, then policy measures directed at helping disadvantaged victims move into safer neighborhoods would be worthwhile.

Footnotes

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.

Notes

ORCID iD

Alyssa M. Yetter

Author Biography

Alyssa M. Yetter is a doctoral candidate in sociology in the Department of Sociology and Criminology at The Pennsylvania State University. Her research interests are in stratification, poverty, and crime. She holds a BA in sociology and anthropology from Denison University, and a MA in criminology from The Pennsylvania State University. She has published in Social Networks.

References

Astone

N. M.

McLanahan

S. S.

(1994). Family structure, residential mobility, and school dropout: A research note. Demography, 31, 575-584.

Baker

C. K.

Cook

S. L.

Norris

F. H.

(2003). Domestic violence and housing problems: A contextual analysis of women’s help-seeking, received informal support, and formal system response. Violence Against Women, 9, 754-783.

Bromet

Sonnega

Kessler

R. C.

(1998). Risk factors for DSM-III-R posttraumatic stress disorder: Findings from the national comorbidity survey. American Journal of Epidemiology, 147, 353-361.

Bures

R. M.

(2003). Childhood residential stability and health at midlife. American Journal of Public Health, 93, 1144-1148.

Burgess

A.W.

Holmstrom

L.L.

(1979). Adaptive strategies and recovery from rape. American Journal of Psychiatry, 136, 1278-1282.

Bursik

R. J.

Jr. (1988). Social disorganization and theories of crime and delinquency: Problems and prospects. Criminology, 26, 519-551.

Dugan

(1999). The effect of criminal victimization on a household’s moving decision. Criminology, 37, 903-930.

Elklit

Shevlin

(2009). Sexual victimization as a risk factor for residential mobility: A case-control study using the Danish registry system. Public Health, 123, 502-505.

Harding

D. J.

Morenoff

J. D.

Herbert

C. W.

(2013). Home is hard to find: Neighborhoods, institutions, and the residential trajectories of returning prisoners. The Annals of the American Academy of Political and Social Science, 647, 214-236.

10.

Herbert

C. W.

Morenoff

J. D.

Harding

D. J.

(2015). Homelessness and housing insecurity among former prisoners. RSF: The Russell Sage Foundation Journal of the Social Sciences, 1, 44-79.

11.

Herzog

H. W.

Jr. Schlottmann

A. M.

(1984). Labor force mobility in the United States: Migration, unemployment, and remigration. International Regional Science Review, 9, 43-58.

12.

Hipp

J. R.

Tita

G. E.

Greenbaum

R. T.

(2009). Drive-bys and trade-ups: Examining the directionality of the crime and residential instability relationship. Social Forces, 87, 1778-1812.

13.

Kilpatrick

D. G.

Acierno

Resnick

H. S.

Saunders

B. E.

Best

C. L.

(1997). A 2-year longitudinal analysis of the relationships between violent assault and substance use in women. Journal of Consulting and Clinical Psychology, 65, 834-847.

14.

Kubrin

C. E.

Stucky

T. D.

Krohn

M. D.

(2009). Researching theories of crime and deviance. New York, NY: Oxford University Press.

15.

Kushel

M. B.

Gupta

Gee

Haas

J. S.

(2006). Housing instability and food insecurity as barriers to health care among low-income Americans. Journal of General Internal Medicine, 21, 71-77.

16.

Logan

J. R.

Alba

R. D.

(1993). Locational returns to human capital: Minority access to suburban community resources. Demography, 30, 243-268.

17.

Long

L. H.

(1972). The influence of number and ages of children on residential mobility. Demography, 9, 371-382.

18.

Macmillan

(2001). Violence and the life course: The consequences of victimization for personal and social development. Annual Review of Sociology, 27, 1-22.

19.

McNeeley

Stutzenberger

(2013). Victimization, risk perception, and the desire to move. Victims & Offenders, 8, 446-464.

20.

Miller

T. R.

Cohen

M. A.

Rossman

S. B.

(1993). Victim costs of violent crime and resulting injuries. Health Affairs, 12, 186-197.

21.

Morenoff

J. D.

Sampson

R. J.

(1997). Violent crime and the spatial dynamics of neighborhood transition: Chicago, 1970-1990. Social Forces, 76, 31-64.

22.

Najdowski

C. J.

Ullman

S. E.

(2009). Prospective effects of sexual victimization on PTSD and problem drinking. Addictive Behaviors, 34, 965-968.

23.

Newman

S. J.

Duncan

G. J.

(1979). Residential problems, dissatisfaction, and mobility. Journal of the American Planning Association, 45, 154-166.

24.

Pavao

Alvarez

Baumrind

Induni

Kimerling

(2007). Intimate partner violence and housing instability. American Journal of Preventative Medicine, 32, 143-146.

25.

Roberts

Horney

(2010). The life event calendar method in criminological research. In Piquero

A. R.

Weisburd

(Eds.), Handbook of quantitative criminology (pp. 289-312). New York, NY: Springer.

26.

Schachter

J. P.

(2004). Geographical mobility: 2002 to 2003 (Current Population Reports No. 20-549). Washington, DC: U.S. Census Bureau.

27.

Shaw

C. R.

McKay

H. D.

(1942). Juvenile delinquency and urban areas. Chicago, IL: University of Chicago Press.

28.

Slocum

L.A.

Simpson

S.S.

Smith

D.A.

(2005). Strained lives and crime: Examining intra-individual variation in strain and offending in a sample of incarcerated women. Criminology, 43, 1067-1110.

29.

South

S. J.

Crowder

K. D.

(1997). Escaping distressed neighborhoods: Individual, community, and metropolitan influences. American Journal of Sociology, 102, 1040-1084.

30.

South

S. J.

Deane

G. D.

(1993). Race and residential mobility: Individual determinants and structural constraints. Social Forces, 72, 147-167.

31.

South

S. J.

Messner

S. F.

(2000). Crime and demography: Multiple linkages, reciprocal relations. Annual Review of Sociology, 26, 83-106.

32.

Speare

Jr. (1974). Residential satisfaction as an intervening variable in residential mobility. Demography, 11, 173-188.

33.

Stark

(1987). Deviant places: A theory of the ecology of crime. Criminology, 25, 893-909.

34.

Sweeten

(2012). Scaling criminal offending. Journal of Quantitative Criminology, 28, 533-557.

35.

Waltermaurer

McNutt

L.-A.

Mattingly

M. J.

(2006). Examining the effect of residential change on intimate partner violence risk. Journal of Epidemiology and Community Health, 60, 923-927.

36.

Western

Braga

A. A.

Davis

Sirois

(2015). Stress and hardship after prison. American Journal of Sociology, 120, 1512-1547.

37.

Xie

McDowall

(2008a). The effects of residential turnover on household victimization. Criminology, 46, 539-575.

38.

Xie

McDowall

(2008b). Escaping crime: The effects of direct and indirect victimization on moving. Criminology, 46, 809-840.

39.

Zweig

J. M.

Crockett

L. J.

Sayer

Vicary

J. R.

(1999). A longitudinal examination of the consequences of sexual victimization for rural young adult women. The Journal of Sex Research, 36, 396-409.