3.2 Measures
3.2.1 Assessment of stalking victimization (Studies I-III)
A stalking questionnaire by Sheridan, Davies and Boon (2001) was used, with minor changes.
The questionnaire as a whole measures participants’ perceptions of what constitutes stalking and their personal experiences of stalking. This instrument was chosen because Finland is one of the countries taking part in an international study on stalking and all data is collected via the measure presented in this study. This instrument has been tested and employed in several studies (Jagessar & Sheridan, 2004; Sheridan, Davies & Boon, 2001; Sheridan, Gillett &
Davies, 2000, 2002). In the first section of the questionnaire the respondents were asked to choose what they considered to be unwanted behavior from a list of 47 behaviors. In the second section the respondents were asked if they had personally been subjected to these 47 behaviors. In the third section the participants were first provided with the following working definition of stalking: “Persistent unwanted behavior consisting of several attempts to approach, contact, or communicate that the recipient didn’t want and did not encourage”.
This definition was followed by a question “Have you ever experienced persistent, unwanted attention (stalking) from a man/woman?” The three response categories “Never/Yes, one time during my lifetime/Yes, several times during my lifetime” were used to determine two subgroups of victims and one of non-victims. If they had answered “Yes” to the previous question, respondents were asked to indicate which of the 47 stalking behaviors they had experienced and how many times. If a participant had experienced several stalking episodes the respondent was asked to answer with reference to the most recent experience only.
Measures and central variables derived from the above-mentioned stalking questionnaire for purposes of the present studies were (Studies I-III):
Stalking was measured by a question “Have you ever experienced persistent, unwanted attention (stalking) from a man/woman?” The three response categories “Never/Yes, one time during my lifetime/Yes, several times during my lifetime” were used to determine two subgroups of stalking victimization and two subgroups of stalking episodes.
Participants who answered “Yes” to the stalking question (as described above) were classified as victims of stalking for further analysis, including incidence rates.
40
Participants who answered “Yes, once during my lifetime” or “Yes, several times during my lifetime” were divided into victims who had experienced one episode of stalking and those who had experienced two or more episodes of stalking respectively. In the studies this variable is referred to as the number of stalking episodes.
Stalkers were categorized into strangers, ex-partners (ex-boy/girlfriend/spouse) and acquaintances (friend/relative/acquaintance/work or study colleague), in line with previous research (e.g. Hall, 1998; Meloy 1998; Sheridan & Davies, 2001). In the studies this variable is referred to as victim-stalker relationship.
The participants were asked to indicate the duration of stalking in years/months/weeks/days, after which the duration was summed up and measured in days for data analysis.
Stalking violence was measured by summing the eight variables consisting of violent stalking behaviors and threats (hurting physically, forced sex, behaving aggressively when seeing victim with other people, confining the victim against his/her will, hurting third parties, threatening to kill, threatening to hurt physically, threatening to kill him/herself) (Cronbach’s alpha = 0.77). The victims were classified as victims of violent stalking if they had indicated experiencing any of these eight behaviors in section three.
3.2.2 Assessment of coping behaviors (Study III)
Coping behavior was assessed by means of a checklist consisting of 35 items. The respondents were asked to check each coping behavior (one or more) they had used in response to stalking. The items were drawn from previous stalking literature concentrating on students’ coping responses. Spitzberg’s (2002) and Spitzberg’s and Cupach’s (2003) coping tactic categories served as a conceptual base for the questions presented to ensure that a broad range of behavioral coping tactics was presented.
3.2.3 Assessment of coping strategies (Study III)
The Ways of Coping Questionnaire – revised (WCQ-R) (Folkman & Lazarus, 1985) was employed to determine how the respondents coped (problem and emotion-focused coping) with being stalked. The WCQ-R is one of the most widely used coping scales and has been utilized in a variety of contexts. The WCQ-R is a 66-item questionnaire that uses a four-point Likert scale designed to assess coping with stressful encounters. Participants responded by rating each coping strategy according to the degree of use: “not used”, “used somewhat”,
41
“used quite a bit”, and, “used a great deal”. The subscales (factors) represent eight ways of coping: confrontive coping, distancing, self-controlling, seeking social support, accepting responsibility, escape-avoidance, planful problem solving, and positive reappraisal. The scales ‘planful problem solving’ and ‘confrontive coping’ are described as efforts to alter the situation and are therefore characterized as problem-focused coping. Furthermore,
‘distancing’, ‘self-controlling’, ‘accepting responsibility’ and ‘escape-avoidance’ are described as emotion-focused coping. ‘Seeking social support’ and ‘positive reappraisal’ are more mixed in character and could be seen as either, depending on the situation.
3.2.4 Assessments of violence victimization (Study IV)
A set of 13 items concerning violence victimization was sent to a sample of university students who had earlier participated in the Student Health Survey 2004 (see Kunttu &
Huttunen, 2005) by responding to a comprehensive range of items concerning students’
general, mental, sexual, and dental health, health behavior, use of health care services, social relations, etc. Of these Student Health Survey items, a set of items regarding mental health symptoms, psychological symptoms (GHQ-12), general health, abdominal symptoms, alcohol consumption (AUDIT), and a symptom index along with the use of health care services were selected as measures for the present study in order to examine the associations to violence victimization. These measures are explained below in more detail.
This additional survey consisted of a set of 13 items regarding violence and threats. The items were selected from the Finnish National Crime Victimization Surveys, being in line with corresponding surveys (e.g. UK, USA) along with violence literature. The respondents were asked if they had ever experienced physical violence or threats. The following working definition of violence was provided on the questionnaire; “Physical violence can occur both within the family and outside the family. The term 'physical violence' here refers to e.g.
pushing, hitting, kicking or sexual violence. The term 'threatening' here refers to any threats directed at you to hurt you physically or damage property belonging to you or your family/friends. This definition was followed by three questions: 1) Have you ever experienced physical violence? 2) Have you ever experienced threats? 3) Have you ever witnessed violence in your family? If answering “Yes” to violence and/or threats, the respondents were asked to estimate how many times they had been exposed to each form of violent experience. The present study aimed to focus on the first two questions, i.e. on the experience of violence and threats. Thus, if answering “yes” to violence and/or threats, the respondents were asked to
42
check all the items that apply from a detailed list of 13 additional items relating to different forms of violence victimization (restraint/pushing/shoving, thrown objects at, attempted hitting or kicking, hitting or kicking, attempted strangulation, weapon violence, attempted sexual violence, sexual violence, other violence, threatened with bodily harm, third party threatened with bodily harm, threatened with property damage, other threats). A separate variable for violence victimization was created by summing up these 13 variables (Cronbach’s alpha = 0.81). For the purpose of this study, the respondents were classified as victims of violence if they had indicated experiencing any of the 13 behaviors mentioned above. The respondents were also asked to provide information about the relationship with the offender by checking a list of eight relationship items (stranger, intimate/ex-intimate partner, acquaintance/studies, acquaintance/work, friend, parent, sister/brother, relative, other), which were coded as separate dichotomous variables for further analysis. Lastly, information was requested regarding physical injury, single vs. multiple incident(s), and the timeframe of the violent experience(s) (onset of violence, latest violent experience, and present situation concerning violence).
3.2.5 Assessment of health problems and alcohol consumption (Study IV) 3.2.5.1 General and mental health symptoms
A 26-item set of questions was used to obtain information on the respondents’ general health.
The questions concerned general health, musculoskeletal, abdominal, cardiovascular, respiratory, dermatological, and dental health symptoms. The respondents were asked to indicate how often they had experienced each of the presented symptoms (Never/occasionally/weekly/daily or almost daily). Symptoms endorsed were then summed to sum variables. The subscales selected for further analysis were: symptom index (all symptoms), general symptoms (headache, dizziness, tiredness, lack of energy/power), abdominal symptoms (abdominal pain/heartburn, nausea/vomiting, gas pains/swelling, constipation/diarrhea) and, mental health symptoms (sleeping problems, problems in maintaining attention/focus, tension, nervousness, depressive symptoms, anxiety) (Kunttu &
Huttunen, 2005).
3.2.5.2 Psychological health problems
The 12- item General Health Questionnaire (GHQ-12) was used to obtain an overall index of the participants’ rating of their psychological distress. The GHQ-12 measures non-specific
43
psychiatric morbidity with 12 items concerning common psychological symptoms such as loss of confidence or sleep disturbances, and symptoms related to anxiety and depression.
The respondent was asked to rate on a 4 –point Likert scale the extent to which he/she has recently experienced 12 specific symptoms in which 1=not at all, 2=same as usual, 3=somewhat more than usual, and 4=much more than usual. The GHQ-12 allows for the derivation of an index of a "case" criterion, i.e. a cutoff for a level of clinically significant pathology. In the present study, if the respondents rated 3 or 4 in four or more items it was considered as an indication of psychological distress, as in the studies validating the GHQ-12 (Goldberg et al. 1997) and prior studies in the Finnish population (Holi, Marttunen &
Ahlberg, 2003; Penninkilampi-Kerola, Miettunen & Ebeling, 2006). The reliability of the scale was .89 for the present sample, assessed by Cronbach's alpha. For the purpose of this study, the scores were summed up to a total GHQ-12 score (psychological symptoms) and used as a continuous variable in data analysis.
3.2.5.3 Alcohol use
The AUDIT test (Alcohol Use Disorders Identification Test) was used to identify persons with hazardous and harmful patterns of alcohol consumption. On the basis of experience gained from the use of AUDIT in research, the following interpretation of the AUDIT scores is suggested: Scores under the cut-off of 8 (the range of 0-7) represent a low level of alcohol-related problems, scores between 8 and 15 represent a medium level of alcohol problems, scores of 16 and above represent a high level of alcohol problems (Babor, Higgins-Biddle, Saunders & Monteiro, 2001, AUDIT-manual, WHO). Thus, the same cut-off points were also used in the present study. The AUDIT has been used in several prior studies in Finland along with population based studies (e.g. Aalto, Alho, Halme & Seppä, 2009; Holmila, 1995). The reliability of the scale was .81, assessed by Cronbach's alpha. For the purposes of this study the individual AUDIT scores were summed up to a total score and used as a continuous variable in data analysis. Also some analysis was run using the audit scores as a dichotomous variable (cut-off point ! 8)
3.2.6 Assessment of health care usage (Study IV)
The respondents were asked whether they had used student health care services during past 12 months. If answering “yes” the respondents were asked to report how many times they had used each service (0= did not use/1=once/2= 2-5 times/3=over 5 times). The nurse’s
44
appointment variable included only visits to a nurse and the doctor’s appointment variable included visits to both general practitioners and specialists (excluding psychiatry).
Furthermore, the mental service variable included visits to both psychologists and psychiatrists. Frequent health care usage was measured by summing the variables consisting of frequent visits (over 5 times during the 12 months) to nurse’s, doctor’s or mental services.
The respondents were classified as frequent health care users if they had indicated frequent visits to any of these three health care services.
3.3 Statistical analysis
The statistical analyses were conducted using SPSS for Windows (version 15 and 16) in all studies (I – IV), and Mplus (version 4.2) in Study II. In Study I, logistic regression was used for predicting violent stalking. Furthermore, two separate logistic regression analyses were conducted to find the best predictors for two forms of violent stalking: physical and sexual violence. The selection of stalking behavior variables for the preliminary regression models were based on chi-square tests along with the body of research on stalking violence.
In Study II, the stalking behavior variables were entered to exploratory factor analysis using the WLSMV (weighted least squares mean and variance adjusted) estimator to uncover the underlying structure and find the optimal number of dimensions. The final three-factor solution was estimated using the maximum likelihood (ML) method in order to maximize the reliability of the factor scores. In the following phase of analysis, a mixture analysis (Mplus) was used to find latent subgroups where the profiles differ from each other regarding the stalking dimensions. This type of mixture model is referred to as ‘latent profile analysis’
(LPA) in the literature. Finally, the duration of stalking was predicted by entering the stalking dimensions and subgroups discovered in prior analysis, the victim-stalker relationship variable, along with some other background variables into a general linear model (GLM).
In Study III, a multivariate analysis of variance was computed using the general linear model procedure (MANOVA GLM) to determine whether differences existed in the use of the eight WCQ coping strategies (dependent variables) regarding seven theoretically central independent variables. Those independent variables, which emerged as individually significant, were selected for further analysis and entered to a second MANOVA. Subsequent univariate ANOVAs revealed statistically significant differences on the WCQ-R subscales.
45
Study IV examined whether violence (victims/non-victims) and gender have any effect on symptoms via Anova GLM using a two-way between-subjects design followed by separate univariate analysis for each symptom category. Also, in order to analyze the effect of the victim-offender relationship on victim symptoms, a series of Anovas (Anova GLM) were conducted. All the findings were considered significant if p < .05. However, the Bonferroni correction was used when appropriate to avoid possible false significant results due to multiple significance tests.
46
4 Results
4.1 Prevalence of stalking among Finnish university students (Study I)
4.1.1 The prevalence and nature of stalking
Altogether 48.5% (n = 298) of the respondents reported having been stalked; 22.3% had been exposed to one episode of stalking (i.e. one period of stalking as defined in the stalking questionnaire explained in the materials section) and 26.2% to two or more episodes of stalking (i.e. several periods of stalking) in their lifetime. The victims had experienced on average 10 different types of stalking of the 47 stalking behaviors. The most common stalking behaviors reported by the victims were: receiving unwanted letters, notes or e-mails (56.4%), receiving multiple, unwanted telephone calls (52.7%), and being verbally abused (50.7%) (see Study I, Table 1 for more detailed information on frequencies). As a whole, stalking was significantly more frequent among female students compared to male students (52.4% vs.
23.2, p < .01).
4.1.2 Victim and stalker characteristics
The majority of stalking victims (N = 277, 93.6%) were women, whereas most of the stalkers were male (91.8%). The average age of the victims was very much in line with the whole sample, i.e. 24.1 years (SD = 3.60, range 18-41 years). The average age of the stalkers as estimated by the respondents was 26.6 years (SD = 8.8, M = 25 years, range 10-60 years). Of the stalkers, 9.7% were 18 years or younger at the time of the stalking, 59.7% were 19-30 years old and 16.4% were over 30 years. In total, 24.8% of the stalkers were ex-partners, 55.0% were acquaintances (e.g. fellow students, co-workers, clients), and 18.5% were strangers.
The mean duration of stalking was 301 days (SD = 488, Md = 90). The stalking duration was significantly associated with the stalker gender (Mann-Whitney U = 2567, p < .05) and the prior victim-stalker relationship (Kruskall Wallis !2 = 45.699, p < .001). Female stalkers were more persistent (M = 572.2, SD = 797.1) compared to male stalkers (M = 273.0, SD = 440.3,) with the size of the gender effect being relatively large (Cohen’s d = .61). The average duration was significantly shorter among strangers (M = 60.4 days, SD = 102.7), than among
47
ex-partners (M = 343.4 days, SD = 429.4) or acquaintances (M = 360.2 days, SD = 564.5).
The size of this relationship effect was also relatively large (Cohen’s d = .60). Table 1 shows the distribution of the duration with regard to the stalker-victim relationship.
Table 1. The distribution of stalking duration in days with regard to the stalker-victim relationship.
Days Stranger% (n=52) Acquaintance% (n=151) Ex-partner% (n=72)
1-21 61.5 23.8 6.9 course of stalking. The number of violent behaviors experienced was significantly associated with the victim-stalker relationship (!2 = 15.752, p < .001); victims of ex-partners reported on average more violent stalking behaviors (M = 2.05, SD = 2.11, range 0 - 8) than acquaintances (M = .67, SD = 1.08, range 0 - 6) and strangers (M = .45, SD = .86, range 0 - 4). Differences in the use of stalking violence between the relationship groups can be seen in Table 2. It is also of note, that the number of violent behaviors was not related with stalker age or gender.
Table 2. The distribution of violent stalking behaviors and the victim-stalker relationship Stranger
48
Two separate logistic regression analyses were conducted to find the best predictors for two forms of violent stalking: one for physical violence and one for sexual violence. In the first analysis, 17 stalking behavior variables based on chi-square tests along with the body of research on stalking violence were entered into the regression model for physical violence (for more detailed information see Study I results section and Table 4). In the final model, three variables produced an overall significant model (!2 = 92.491, p < .001) with an R square of 0.48, and with a predictive accuracy rate of 91.6% (Hosmer-Lemeshow test !2 = .22, p = .90). The three predictors for physical violence were: death threats (p < .01, OR = 7.47), property damage (p < .001, OR = 8.79) and confining the victim against his/her will (p < .001, OR = 9.41).
In the second regression analysis, the analysis for sexual violence, 15 stalking behavior variables were entered into the regression model (for detailed information see Study I results section and Table 5). In the final model three variables, i.e. verbal abuse (p < .01, OR = 6.30), threats to hurt physically (p < .001, OR = 5.82) and making arrangements without asking the victim first (p < .01, OR = 4.67), accounted for 28% of the variance in sexual violence and produced an overall significant model (!2 = 37.57, p < .001) with a predictive accuracy rate of 93% (Hosmer-Lemeshow test !2 = 1.00, p = .80). It is of note that the inclusion of demographic variables (age and gender of the stalker, relationship with the stalker), stalking duration, and the number of stalking episodes did not emerge as significant predictors for either physical or sexual violence.
4.2 Latent profile analysis approach to duration of stalking (Study II)
Study II consisted of a subsample of 137 respondents. This subsample included only those students who had been subjected to one stalking episode during their lifetime (22.3% of the whole sample), and was selected/created in order to analyze individual stalking behaviors (see methods section for sample characteristics). Stalker characteristics and the duration of stalking in this subsample were similar to the whole sample as reported in substudy I.
4.2.1 Defining the dimensions of stalking behavior
The present study was interested in finding the optimal number of stalking behavior dimensions underlying the stalking behavior variables presented in the current study. The factor analysis resulted in a three-factor solution and three distinct dimensions of stalking
49
behavior could be interpreted: violence, surveillance and contact seeking (see Table 1). The violent dimension includes stalking behaviors, which comprise intrusive, threatening and/or violent behaviors. The surveillance dimension includes behaviors, which can be characterized as ‘classic’ stalking behavior; including following and loitering near locations where the stalker might see or run into the victim. The third dimension was named contact seeking as it includes stalking behaviors indicative of attempts to learn more about the victim, to get to know him/her, and possibly develop a relationship.
Table 1. A three-factor solution for stalking behavior with items ordered according to the difficulty parameter
Dimension 1 Violence Dimension 2 Surveillance Dimension 3 Contact seeking
Secretly taking belongings Seeing a person at the same time each day Changing school etc. to be near the victim Intercepting mail Trying to get to know victim’s friends Trespassing on victim’s property Forced sexual contact Changing school etc. to be near the victim Sending or giving gifts
Physical threats Spying Offering to buy a drink in a café or bar
Death threats Loitering by victim’s home Engaging in a conversation in a public place
Death threats Loitering by victim’s home Engaging in a conversation in a public place