Who is Transitioning out of Prison? Characterising Female Offenders and Their Needs in Chile

The last decades’ increase in female incarceration has translated into an increasing number of women being released from prison. Understanding their characteristics and criminal trajectories can enlighten us regarding the different needs of women upon re-entering society after incarceration. Drawing on data from the Reinserción, Desistimiento y Reincidencia en Mujeres Privadas de Libertad en Chile study, this article identifies different profiles among a cohort of 225 women who were released from prison in Santiago, Chile, and demonstrates that significant heterogeneity exists among them in terms of their criminal trajectories and the intervention needs to support their transition out of prison.


Introduction
The increase in the population of incarcerated people over the last few decades has subsequently increased the number of people who are released from prison. There is consequently a salient interest in the re-entry process. Chile, like other Western countries, has experienced a significant rise in its prison population since the 1990s (Salinero 2012), with a relatively larger increase being driven by the female population (Gendarmeria 2016;Walmsley 2017).
Although most of our current understanding about offending is based on research conducted among male offenders, a substantial body of research exists that, from the feminist perspective, analyses the distinctive criminal trajectories of female offending and the specific needs of women re-entering society after being incarcerated (Belknap 2007;Brennan et al. 2012;Daly 1992;Salisbury et al. 2017;Simpson et al. 2008). Nevertheless, most of this research has been conducted in anglophone countries, which neglects both the specific historical and societal issues from the global south and the particular ways in which crime has been affected by culture and gender.
In the present study, we identify and characterise different profiles among a cohort of women who have been released from prison in Santiago, Chile's capital city. We commence by discussing the evidence regarding female criminal trajectories. Then, based on a set of indicators that have been derived from the literature, we use latent class analysis (LCA) to identify different groups among this cohort of women offenders. After describing each group in terms of sociodemographic variables and life histories of offending and victimisation, we discuss policy suggestions for addressing the differences between the identified groups.

Women's Criminal Trajectories
Research on criminological careers has revealed that women desist from crime earlier than men (Flood-Page et al. 2000;Giordano et al. 2002;McIvor, Murray and Jamieson 2004) and that they have much lower re-offending rates (Cannon and Wilson 2005; Ministry of Justice 2011). Compared to men, women released from prison have, on average, shorter criminal histories that start later in life and are less likely to have been sentenced for serious and violent crimes (Block et al. 2010;Morales et al. 2015). However, these women leave with several disadvantages as they transition from prison back into their communities. Incarcerated women are more likely than incarcerated men to have lower levels of education and histories of unemployment (Kruttschnitt and Gartner 2003;Morales et al. 2015). They are also more likely to report mental health problems and problematic drug use (Bloom, Owen and Covington 2003;CESC 2007;Hurtado and Larroulet 2012), as well as to have experienced histories of victimisation and abuse during childhood from their family members and as adults from their romantic partners (Belknap 2007;Britton 2011;Larroulet 2016).
Since the early 1990s, feminist scholars have developed what is now termed 'gendered pathways'-an approach that analyses the life experiences that have led women and girls into offending (Miller and Mullins 2006). The main focus of inquiry is to explain the nature of female pathways into law-breaking, with an emphasis on biographical elements (Daly 2013).
Although this line of inquiry emerged early in feminist criminology (for a review, see Belknap 2007), one of the most cited studies on gendered pathways to crime is Daly's work, in which she analysed the files of 40 women and 40 men to determine gender differences and similarities regarding the pathways to felony court in the United States (US) (Daly 1992(Daly , 1994. Among females, Daly (1992) identified five different pathways that led women into the criminal justice system. The more common pathways included street women and the harmed and harming women. Street women were characterised by an early abandonment from an abusive home that pushed them into a life on the streets-one marked by early involvement in offending and high levels of IJCJ&SD 114 www.crimejusticejournal.com © 2020 9(1) substance use and abuse. Harmed and harming women comprised women who were victimised as children and who were characterised by aggressive behaviour that appeared early in life and that extended into adulthood, which linked to mental health problems and problematic drug use. A third pattern that Daly identified included drug-connected women who became involved in offending through the drug market, mainly as a result of their relationship with family members or romantic partners who were either using or selling drugs. A shorter involvement in offending characterised this pattern, as well as an almost non-existent use of drugs. Finally, there were two patterns that were characterised by a later crime onset. The first group included the battered women, which consisted of women who became involved in crime due to their relationship with a violent partner (see also Richie 2001;Simpson et al. 2008). The final group, called the economicrelated pathway, involved women who exhibited lower levels of risk factors that traditionally linked to offending and who had committed mostly economic offences, such as fraud (Daly 1992).
Subsequent studies have confirmed the existence of similar patterns, mainly highlighting the role that domestic violence and economic and social vulnerability plays in women's crime involvement (Brennan et al. 2012;Nuytiens and Christiaens 2017;Richie 2001;Simpson et al. 2008). Research has also evidenced the influence of drug abuse on female crime, highlighting the emotional stability acquired by women who successfully began avoiding drug use and demonstrating how this was crucial in their abandonment of crime (Barry 2012;McIvor et al. 2009;Michalsen 2019;Rumgay 2004).
However, most of this research has been conducted in anglophone countries (for one exception, see Nuytiens and Christiaens 2016). Feminist criminology in Latin America has mainly explored the prison experiences of incarcerated women (see Azaola 2013;Lemgruber 1999;Salazar and Cabral 2012), and less attention has been paid to their criminal trajectories and to the heterogeneity that might exist between different groups of women. Nevertheless, some studies have confirmed the roles that economic marginalisation and gender violence play as fundamental issues that contribute to women's offending, especially in the case of drug-related crimes (Antony 2007;Boiteux 2015;CELS 2011;Rodríguez 2005). In a recent study conducted with women who were convicted in a federal jail in Argentina, Salisbury et al. (2017) highlighted the high prevalence of victimisation and violence that was evident before criminal involvement, as well as the economic motivation underlying most of the crimes. Regarding gendered pathways, most of the women in federal prisons in Argentina fell into the drug-connected and economic-motivated pathways, while fewer women reported life histories of severe abuse and victimisation.
Research has overall revealed that incarcerated women are people with life stories, backgrounds and circumstances that have been embedded into structural contexts characterised by substantial gender inequalities-all of which must be considered when developing re-entry policies (Chesney-Lind 1997;Gelsthorpe 2004). A better understanding of the diversity that underpins female offenders' criminal trajectories may help us develop programs for those women transitioning out of prison, especially in Latin America, as there is scarce information regarding women offenders. Accordingly, this article explores the characteristics evidenced in a sample of women who were released from prison in Santiago, Chile. It does so to assess whether heterogeneity is present among the women, and, thus, to contribute with the development of interventions that better account for the specific needs of this population.

Data
In the present study, we used data from the Reinserción, Desistimiento y Reincidencia en Mujeres Privadas de Libertad en Chile study ('reintegration, desistance and recidivism among female inmates in Chile'), an intensive longitudinal study that followed a cohort of women who were The target population was Chilean women who served custodial sentences of at least 30 days. Two hundred and seventy-seven women met the selection criteria during the observation period, of which 225 (81 per cent) participated in the study, 22 (eight per cent) refused to participate and 30 (11 per cent) where not contacted due to logistic issues. 2 The study consisted of a baseline interview that was conducted one to two weeks before prison release and four follow-up interviews between a week and 12 months after release. The current analysis is based on the information collected in the baseline interview, which mainly included closed-ended survey questions regarding life histories in different domains (e.g., work and family histories, criminal trajectories, drug use and personal relationships) and the women's experience in prison. The recruitment and interviews were conducted by a team of trained social workers, and the women respondents received an economic incentive for their participation. 3 Regarding the specific context, Santiago, like other Latin American capitals, concentrates a significant share of the country's total population, concentrating approximately 30 per cent of the female prison population. 4 Moreover, high levels of urban segregation characterise the city, which results in unequal access to education, other services and labour market (Fuentes et al. 2017). As for the rest of Chile, females are more affected by poverty and have poorer participation in the labour market. 5 While the UK is ranked 18th and the US 28th out of 145 countries in the Gender Gap Index, Chile is ranked 73rd (World Economic Forum 2018).

Analytic Plan
We use LCA to identify different profiles among the released women (Collins and Lanza 2010). The premise of the latent class model is that the observed heterogeneity in the sample is the consequence of a discrete set of unobserved groups from which the population is drawn.
As noted above, the groups were defined based on variables that tied to the literature regarding female criminal trajectories, which were available on the data set. 6 Specifically, we used seven binary indicators, three of which directly related to the type of crime in which women had been involved throughout their lives: whether they had ever been involved in 1) theft (70 per cent); 2) other property offences, including violent offences such as robbery (59 per cent); and 3) drug offences (44 per cent). 7 Three indicators regarding the level of criminal involvement were also included: 4) early onset in offending (37 per cent), which was assessed by having committed the first offence at age 14 or younger; 5) self-identification as a delinquent (22 per cent); and 6) prior prison sentences (26 per cent), which were measured as being above the median in the sample distribution (corresponding to four or more sentences). Finally, we included a measure of whether the woman presented 7) symptoms of drug abuse or drug dependency at a baseline level (40 per cent). Table 1 outlines the descriptive statistics of these indicators. The mean corresponds to the proportion of women who show positive values for each one of the binary indicators.
We selected the model that fitted the data better and evaluated it in terms of the later classification of group membership. We then classified each woman by the group in which she had a more substantial likelihood of membership. Each of those groups was then described in terms of socio-demographic factors and life histories of offending and victimisation to address the question regarding their specific needs and to propose tailored policies for the groups.
Significant differences between groups were tested using an ANOVA test.

Results
The first step in the LCA was to run models with a different number of classes and compare them in terms of the fit statistics provided by the models (BIC and AIC 8 ). Although the models with more than one class significantly fit the data better, the differences between the two and threeclass models were not as apparent. 9 After analysing the different indicators, we continued with the three-group solution that provided a more nuanced representation of the population of interest. Moreover, the posterior probabilities of group membership in the three-class model were larger than .70, which is considered a threshold for correct classification (Roeder et al. 1999). 10 Figure 1 displays the item-response conditional probabilities, which represent the probability of answering 'yes' to the given item, conditional on group membership. Class 1 (31 per cent) is characterised by its high prevalence of drug offences and low prevalence of the remaining types of crimes. Women in this class have lower incarceration rates and started offending at older ages. They do not identify themselves as delinquent, nor do they present a high likelihood of drug abuse.
Women from Class 2 (29 per cent) also presented a lower likelihood of deep involvement in criminal behaviour, as well as a lower probability of having experienced many prior incarceration episodes or of presenting symptoms of drug abuse. Although their chances of having an earlier crime onset were higher than those among the women from Class 1, Class 2 women also do not identify themselves as delinquents. The main difference between Classes 1 and 2 is the type of offence in which they have been involved: Class 2 has a lower likelihood of having been involved in drug offences, but a higher likelihood of having been involved in theft and other property crimes than women from Class 1.
Finally, Class 3 is characterised by its higher involvement in offending; women in this class were more likely to report having had four or more prior incarceration terms, having committed their first offence when they were 14 years old or younger and perceiving themselves as delinquents.
They are also significantly more likely to present symptoms of drug abuse and dependency. Women in this class are predominantly involved in thefts and other property offences.

Different Life Histories, Different Classes
To describe the different classes, we grouped each woman according to where she had exhibited a larger likelihood of class membership. As Table 2 reports, the classes differed in terms of sociodemographic characteristics. Women in Class 1 were, on average, 10 years older than the other two classes. Women from Class 2, although younger than those from Class 1, were slightly older than those from Class 3 and had higher chances of having achieved a high school diploma. In terms of offending, 90 per cent of the women classified in Class 1 were sentenced for a drug offence, while 66 per cent of those classified in Class 3 were sentenced for theft. Those sentences were associated with longer and shorter sentence terms, respectively.
Women from Class 2 presented a more complex criminal pattern. Although their current offence was a property crime, they were more likely to have been sentenced for property crimes that involved violence (i.e., robberies). Similar to the women from Class 1, they were more likely to be in their first incarceration term. Although women from Class 1 reported fewer factors that were traditionally associated with a higher likelihood of offending (i.e., an early crime onset, early use of hard drugs or deviant peers), they were as likely as women from Class 2 to report having ever had a parent who was incarcerated.
The three classes differed in their experiences before their incarceration. Women from Class 3 were more likely to experience violence during childhood, life on the streets or in foster care, and they were more likely to have had an earlier involvement in the criminal justice system. Their lives prior to prison were also characterised by residential instability and association with criminally involved peers. Although no significant difference was found in terms of intimate partner violence, women in Class 3 were more likely to report having assaulted their partners  In terms of access to services during imprisonment (Table 3), the differences between classes were also observed in training and employment. Specifically, women from Class 3 were significantly less likely to have had access to employment, training or even drug treatment while IJCJ&SD 119 www.crimejusticejournal.com © 2020 9(1) incarcerated-even though almost 90 per cent of them exhibited symptoms of drug abuse or dependency. Their lack of access to programs is probably explained by the fact that the Chilean corrections system prioritises inmates with larger sentences when offering access to services (see Baldry 2010). Although women from the other groups did have greater access to services, their participation was also reduced, with only 57 per cent of Class 1 and 49 per cent of Class 2 having had access to training and employment interventions.

Discussion
The present study explores the characteristics of a sample of women who were released from prison in Santiago, Chile, to assess whether there are different profiles of incarcerated women and to contribute to the development of tailored interventions that account for the specific characteristics and needs of this population. Three different groups of women were identified in the sample, which can be linked to the specific taxonomies that were presented in previous studies on women's pathways into crime (Brennan et al. 2012;Daly 1992Daly , 1994Simpson et al. 2008) and to the results from other studies on female offenders as well.
Class 1 revealed a pattern that was characterised by its late onset in offending and, most importantly, by its involvement in drug offences. As expected, these women were older and had fewer of the factors that characterised high-risk offenders from a risk-need-responsivity perspective (Andrews and Bonta 2010). The above is not surprising, as other authors have reported that the number of women incarcerated for drug trafficking has significantly increased over the last decades in Latin America (Salisbury et al. 2018;Transnational Institute 2011). According to the most recent data available, the female prison population in Latin America almost doubled from 2006 to 2012, from 40,000 to more than 70,000 inmates, mainly due to drug offences (Youngers 2014). The combination of a socio-economic disadvantage, a precarious employment and easy access to illegal drug markets could partly explain the women's decisions to engage in this type of crime (Larroulet 2011;Salisbury et al. 2018). Indeed, research in Latin America has shown that most women become involved in drug offences because they were driven by poverty or financial needs, in which the drug business was a means for contributing to the family income (Anitua and Picco 2017;del Olmo 1998;Hübschle 2014). In several Latin American countries, the previous economic activities of most of the women involved in the drug trade included domestic labour and prostitution (UN Women 2014). Regarding previous research, it can be argued that this class would be consistent with Daly's drug-related women, who are characterised by a late onset in offending and, importantly, an involvement in drugs offences.
Class 2 suggested a more complex pattern than what has been shown in prior studies. Although women in this class reported a later involvement in offending and less embeddedness in criminal IJCJ&SD 120 www.crimejusticejournal.com © 2020 9(1) networks compared to women in Class 3, they did report higher levels of criminal involvement and unstructured lifestyles than women from Class 1. As opposed to the harm and harming women pattern presented by Daly, women in Class 2 reported having been less exposed to violence, either as a victim or as a witness, when compared to those in Class 3. Some women in this group may be associated with the situational pathway reported by Brennan et al. (2012), which mixed women who had fewer of the risk factors traditionally associated with criminal behaviour, who became mostly involved in minor property or drug offences, who had a late onset in offending and who did not have school or psychological problems. However, other women in this group reported being involved with criminal peers and were only slightly less likely than those in Class 3 to report having carried and used weapons as a minor. They were also overrepresented among those currently sentenced for more severe property offences, such as robbery. These women were just as likely as those from Class 1 to deny the delinquent identity, which might be considered a degrading label by this group of women. This finding might be interpreted as a way these women can portray themselves as a professional thief, in which criminal activity resembles traditional masculine criminal behaviour that allows them to engage in so-called 'cross-gendering' practices. Prior work conducted among young female offenders in Chile is consistent with these findings. Droppelmann (2019) observed that young female offenders persisted in crime as a means of resisting patriarchy and subordination by facilitating financial independence and avoiding the acquisition of roles that were anchored in the ethics of care. This pattern reinforces the claim that, for several women, crime operates as a way of establishing autonomy and defying constraining gender binaries (Carlen 1988;Miller 2002).
Finally, women from Class 3 were characterised by their deep involvement in offending as a lifestyle. These women committed multiple less serious offences and had high recidivism rates, which might correspond to those that previous research has called churners (Leverentz 2019;Lynch and Sabol 2001). Churners are defined as individuals who move in and out of prison. They usually account for a large share of the prison admissions and releases. Women in this class have been sentenced mostly for theft, which relates to their histories of short-term sentences. Consistent with Baldry's argument (2010), these 'short-termers' constitute a significant percentage among the female prison population, and their profiles differ from those that appear in most studies, as based on prison census data (Baldry 2010). Given these women's marked marginalisation, they are related either to the street women or to the harm and harming pathways reported by Daly (1992) in previous research.
Beyond the commonalities, the results of this study reveal that women from each of the three classes have experienced different criminal trajectories and that they had unequal access to programs while incarcerated. Therefore, they present different needs that are not entirely fulfilled, though they could be addressed through specific programs and interventions. For example, considering that women in Class 1 exhibited a lower involvement in criminal lifestyles and were mainly involved in drug offences, their interventions should be concentrated on training and employment upon release. Those interventions must ensure that the types of jobs they access provide economic independence and that they concurrently represent a source of personal fulfillment to become a real alternative to drug trafficking. In contrast, women from Class 2although less vulnerable than those from Class 3-have different needs. Due to their criminal involvement and lifestyles, and to the benefits that they probably gained from crime, desistance would be a great challenge for them. In that sense, these women would likely benefit from interventions that tackle criminal attitudes (e.g., cognitive behavioural therapy), as well as from employment and training opportunities that allow them to gain independence, self-efficacy and self-worth. In the case of this group, it is especially relevant to help these women recognise the existence and accessibility of the opportunities that are available to them beyond the norms of their gender and beyond the traditional female roles available. Finally, Class 3 corresponds to a profile of multiple intervention needs and a history of reduced access to social protection services. Considering their backgrounds of brief prison convictions and the evidence regarding the incremental effects of those short sentences on recidivism (Armstrong and Weaver 2010; The IJCJ&SD 121 www.crimejusticejournal.com © 2020 9(1) Howard League of Prison Reform 2011), alternatives to incarcerations would be a suitable solution for this group. Despite several controversies (see Drugs Security and Democracy Program 2018), drug treatment courts (DTC) have a proven influence on offenders with drug addictions, whose offending is partly related to their drug misuse (Paz Ciudadana and Banco Interamericano de Desarrollo 2018). Beyond DTC, the women of Class 3 could also benefit from several interventions, including cognitive behavioural therapy and mentoring, both before and after imprisonment.

Conclusion
The present study aimed to identify and characterise different profiles among women who were released from prison in Santiago, Chile, as well as to shed light on some of the needs that they would have upon release. Three different classes were identified among the women from the sample. Women from Class 1 were characterised by their primary involvement in drug offences and by a later crime onset. Class 2 includes women who have a later involvement in offending, but who have higher levels of criminal involvement and more unstructured lifestyles than women in Class 1. Finally, women in Class 3 started committing crimes earlier in life, had higher involvement levels in offending (mainly less serious property crimes) and substance abuse, and were more likely to perceive themselves as delinquents than the women from the other two classes. Although these patterns demonstrate similarities with previous research, several interesting findings emerged that clears the path for developing more tailored policy responses for each of these groups, which were discussed above.
In recent years, especially with the adoption of the Bangkok Rules, the inclusion of a genderresponsive approach in penal policies has been widespread and has influenced how prison interventions and programs must be delivered for female offenders (Hannah-Moffat 2011). A 'gender-responsive' approach has been defined as the development of a women-oriented model that 'targets women's pathways to criminality by providing effective interventions that address the intersecting issues of substance abuse, trauma, mental health, and economic marginality' (Bloom et al. 2003: 75). As this article has demonstrated, most of the women being released in Santiago share a history of economic marginalisation and victimisation within the family setting. Nevertheless, despite these similarities, the three groups identified also revealed significant heterogeneity between them in terms of their characteristics and intervention needs. Those differences are not currently addressed in the Chilean prison system. Women can access several services that do not address the factors that would help them leave crime behind. Moreover, as it happens in other countries, women serving short sentences are left behind, without access to services even though they display substantial vulnerabilities and marginalisation.
Although the present article provides new information for improving the knowledge gap regarding women offenders and their re-entry needs in Latin America, some limitations must be considered. First, we must mention that the characteristics of the different groups highlighted in the article offer only a limited insight into the needs that these women will have following their imprisonment. Although some of the factors that lead women to engage in crime might sometimes overlap with their needs after release, several other issues emerge when facing the challenges of re-entry that the present study does not consider. Second, the analysis and the construction of the different classes were limited by the data available in the data set. Therefore, although partly drawing on the female pathways perspective, the data available did not allow us to accurately assess all the key variables that would explain these women's criminal trajectories. Finally, it is essential to consider that several of the women's characteristics in the present analysis are based on deficits, problems and vulnerabilities-several of which are retrospective and static factors that render intervention insufficient. In that sense, interventions should not be limited to addressing those factors only, and correctional practice must move from a deficit-based model to an approach that promotes young women's strengths, decision-making and personal growth. IJCJ&SD 122 www.crimejusticejournal.com © 2020 9(1) IJCJ&SD 123 www.crimejusticejournal.com © 2020 9(1)