Factors associated with spousal physical violence in Albania: cross sectional study

Introduction

Intimate partner violence is present in almost all societies1 2 and is associated with considerable morbidity.1 3 Little is known about the magnitude of intimate partner violence and the factors associated with it in post-communist countries.1 Intimate partner violence is more common in patriarchal societies and settings where violence is commonly used in conflict or to gain ascendance,4 so there has been particular concern about its prevalence in patriarchal transitional countries, where violence has become commonplace and social relations have been disrupted.5

Albania is a small country, and 58% of the population lives in rural areas.6 It has undergone major social and political upheaval in the past decade.5 6 After the collapse of the communist regime in 1990 many people migrated from rural to urban areas.6 7 The population of Tirana, the capital city, increased rapidly from 150 000 in 1991 to more than 500 000 a decade later.6 In the western Balkans there are substantial social and cultural differences between people from urban and rural areas,5 giving rise to an expectation that place of birth (and upbringing) could be an important determinant of attitudes or social norms that might give rise to different behaviour patterns and practices.8 Albania has been described as a particularly patriarchal society.9 10

Intimate partner violence is integrally linked to ideas of male superiority over women.4 These are manifest in different ways in different societies, but violence is usually used to create and enforce gender hierarchy and punish transgressions11^–13; to resolve relationship conflict14; and to seek resolution of crisis of masculinity by providing an (often transient) sense of powerfulness.11^–13 Different factors influence, and indicate, the status of women and men in a society and so influence these processes. These factors include social and demographic characteristics of the women and men, their economic circumstances, and the characteristics of their relationship.15^–20 Other factors, including alcohol consumption, depression, experience of abuse in childhood, and the strength of family and social networks, may additionally influence responses to conflict and a perception of emotional insecurity.1 4 19 20

This study aimed to describe the magnitude of the problem of physical violence among intimate partners in Tirana and the demographic and social factors associated with it.

Methods

We carried out a cross sectional study in Tirana in July-October 2003. At the last census in April 2001, 93 230 women aged 23-63 (at that time) resided in Tirana.7 The sampling frame for this study was the list of these women's names obtained from the National Institute of Statistics. From this list we drew a simple random sample of 1500 women (both married and unmarried), but we included only married women, as cohabiting is unusual in Albania. The study was conducted in 2003 so the women were aged 25-65. Of the 1500 potential respondents, 106 women could not be identified (wrong address or changes of residence), 198 women were excluded as they were not currently married, 129 women were either absent (after two visits), or refused to participate, and 28 questionnaires were incomplete and could not be included in the analysis, giving a response rate of 87% (1039 of 1196).

A structured questionnaire was administered by an interviewer to married women who agreed to participate in the survey after an informed consent form had been signed. The questionnaire included questions on age, education, religion, occupation, place of birth (rural or urban) and age of move to Tirana (if any) of the woman and of her husband. It asked about number of children, household size, and household disposable income (from all sources, after rent is paid). It asked about whether the husband was “violent and constantly jealous,” was drunk (almost) every day, and whether the woman knew if he had been abused by his parents during childhood or had witnessed violence between his parents. Women's recent experience of spousal violence was measured by a question asking whether in the past year she had been “hit, slapped, kicked, or physically hurt” by her husband. Those answering “Yes” were asked how many times this had happened.

We computed income per capita as the ratio of disposable income per family member living in the home (partner, children, parents, inlaws) and categorised as low (< €80 (£54; $95) per capita/month), middle (€80-150 per capita/month), and high (> €150 per capita/month). Educational attainment of women and their husbands was classified as high if they had a university degree (or other post-school qualification), middle if they had a formal education of 9-12 years, and low if formal education lasted no more than eight years. For profession we used the self reported main occupation, categorised as “white collar” (clerk or business occupations), “blue collar” (industry and agriculture), housekeeping (for women), or unemployed or retired.

Data analysis

After calculating descriptive statistics for the prevalence of physical spousal violence and social and demographic variables, we used logistic regression to examine bivariate associations between these variables and Student's t test to compare the mean number of children in a household and number of household members in the two abuse categories. We then constructed two multiple variable logistic regression models. We first modelled factors associated with reporting spousal violence for 945 women who had no missing data, using a backwards stepwise elimination procedure with a P value to exit > 0.10. The model tested the age of the husband, age difference between spouses, place of origin of women, income level, difference in educational level between spouses, difference in employment level between spouses, number of children, number of household members, employment status of men, male jealousy, alcohol misuse, and male experience of abuse in childhood. Colinearity between educational status of men and women and educational discrepancy between spouses prevented all of these being retained in one model. As all seemed to be important in different ways, we built a second model showing educational discrepancy, but not men's and women's education level, otherwise using the same backwards stepwise elimination procedure and candidate variables. The fit of the logistic regression models was tested by likelihood ratio and Hoshmer-Lemeshow goodness of fit tests. All statistical analyses were done with SPSS 10.0 for Windows.

Results

The mean age of the women was 42.3 (median 41.0, interquartile range 33.0-50.0) years, and their husbands' mean age was 46.6 (45.0, 37.0-55.0) years. Of the women interviewed, 384 (37%) reported at least one episode of spousal violence in the past year, and 267 (26%) reported three or more episodes (655 (63%) reported none; 72 (7%) reported one episode, 45 (4%) two, 97 (9%) three, 39 (4%) four, 104 (10%) five to nine, 27 (3%) 10 or more).

The prevalence of violence was highest among women aged 25-34, those with more than 12 years of education, those employed in white collar positions, women who were more educated than their spouse, in households where there were more children and more household members, and in couples where the wife was the only one employed (table 1).

Husbands who had been violent towards their wives were more likely to be aged 25-34, to have less than nine years of schooling, to have come from a rural area (especially doing so at, or after, 10 years of age), and to have witnessed their mothers being abused and to have been beaten in childhood (table 2).

The multiple variable logistic regression model (table 3) of factors associated with physical violence showed greatest risk for the youngest women (aged 25-34). Risk increased with increasing educational level for women, with decreasing educational level for men, and with men being of rural origin and having left rural areas at age 10 or later. Risk of violence reduced with decreasing employment status of women: women in blue collar work, housekeepers, and unemployed women were at lower risk than those in white collar occupations.

The model with educational discrepancy (table 3) showed least risk of violence when women were married to a more educated spouse. Women were protected (but less so) if their spouse had the same educational level, and they were most at risk if they were more educated than their spouse. There was substantial confounding between the educational discrepancy variable and most factors in the first model. The only variable that showed no change in effect size was men's rural origin. After adjustment for education discrepancy, women's age was not independently associated with risk of violence; women's employment status was protective only if women were unemployed; and a variable not in the first model, men's employment status, is seen in the model with increased risk associated with men being unemployed (table 3).

Discussion

Spousal violence is highly prevalent in Tirana, and, indeed, the prevalence of intimate partner violence in the past year is one of the highest reported internationally.1 Women with higher status (those best educated, with higher status employment, and those more educated than their spouses) were most likely to report violence. Less powerful men (those less educated and unemployed) were more likely to be violent. Men who came from rural areas, especially those who had lived in rural areas longer, were more likely to be violent. Younger women were more likely to experience physical abuse

These findings, related to women's and men's status, are in keeping with theories of causes of intimate partner violence that argue that violence is used to enforce gender hierarchies and, particularly, when men have a sense of powerlessness because their social position makes them feel “unsuccessful” as men.4 11^–13 21 Research from other countries has often also shown that the most educationally and socially empowered women gain a level of protection.1 4 19 21 22 Such protection is not seen among the women of Tirana, possibly because Albania is known to be a country with particularly conservative ideas about sex roles.9 10

The finding that men of rural origin were independently at greater risk of abusing women than those of urban origin has not been reported often.4 17 19 The explanation may be cultural or could be due to an unmeasured confounder. In rural areas there may be different norms about a man's right to exert control over his wife, and weaker sanctions against violent behaviour. This might have been supported by evidence that men from rural areas were more likely to have witnessed abuse in childhood, but the level of reporting of this was so much lower than would have been expected from the overall prevalence of spousal violence, and we suspect that the (third party) reporting by women was unreliable.

It is interesting to consider the extent to which these findings have been influenced by the transition in Albania. In the past 15 years, rapid movement of people to cities has occurred, economic inequality has increased, the position of women has changed (albeit slowly), and unemployment has become very high. This has brought rapid changes to many Albanian families, and many women are more independent and have greater economic responsibility whereas men are less able to fulfil their culturally expected roles as (family) protectors and providers.9 10 The tension this has created may have led to an increase in intimate partner violence, especially among the groups of women who have been in the forefront of these changes.

This study has some limitations. Under-reporting of intimate partner violence in research is common, especially when exposure is assessed with one question.23 Sexual violence and emotional violence were not measured. Lifetime exposure was not measured, so we cannot discount the possibility that differences in prevalence between higher and lower status women were due to differences in duration of violence in marriage. Some risk factors for violence, such as alcohol consumption and relationship dynamics or conflict, were not assessed comprehensively, and this may explain why they were not shown to be associated with violence in the final models. Relationships found in one study of this nature cannot be assumed to be causal.