Literature review for research methods

International Journals

DETERMINANTS OF DOMESTIC VIOLENCE: A Cross-National Study Author(s): CHRISTINE ARTHUR and ROGER CLARK Source: International Journal of Sociology of the Family, Vol. 35, No. 2 (Autumn 2009), pp. 147-167 Published by: International Journals Stable URL: https://www.jstor.org/stable/23070721 Accessed: 25-04-2019 20:07 UTC

REFERENCES Linked references are available on JSTOR for this article: https://www.jstor.org/stable/23070721?seq=1&cid=pdf-reference#references_tab_contents You may need to log in to JSTOR to access the linked references.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide

range of content in a trusted digital archive. We use information technology and tools to increase productivity and

facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at

https://about.jstor.org/terms

International Journals is collaborating with JSTOR to digitize, preserve and extend access to International Journal of Sociology of the Family

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms

International Journal of Sociology of the Family, Vol 35, No. 2 (Autumn) 2009

DETERMINANTS OF DOMESTIC VIOLENCE:

A Cross-National Study

CHRISTINE ARTHUR AND ROGER CLARK

Rhode Island College

This paper examines determinants of variation in domestic violence at the national level. Although Levinson (1989) studied domestic violence at the societal level using anthropological data from the Human Relations Area File, most research on domestic violence has been at the individual or household level. This paper examines what determines variation in domestic violence at the national level

using data we gathered from two sources: the United States State Department's (2005) Country Reports on Human Rights Practices and a (1999) Population Reports from Johns Hopkins University. We then tested six theories designed to explain cross-societal variation in domestic violence: a Resource Theory, an Exchange Theory, a Culture of Violence Theory, a Patriarchal Theory, a Modernization Theory and an Economic Dependency Theory. All six theories were supported to some degree by our analysis. However, we did find reason to believe that economic dependency and modernization may be underlying factors for the other determinants we identified.

Domestic violence can involve a wide array of physical, sexual and emotional abuse. Until recently, in most parts of the world, a husband a legal right to his

wife's body. Many countries now provide legal protections against domestic violence, but many do not. Domestic abuse is the leading cause of injury to women even in the world's most "developed" countries, like the United States (Brush 1990), and it has mostly been studied in the United States and other Western nations as an individual- and family-level problem. But there is international variation in the levels of domestic violence (Heise et al., 1999), and understanding this variation is important for a complete understanding of the phenomenon of domestic violence generally.

What we do first in this paper is review some of the literature that looks at the causes of domestic violence at the individual and family level. We then introduce theories that have been developed to explain domestic violence at the societal level, as well as two societal-level theories of our own. In this paper we test these societal-level theories with unique data sets about nation

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms

148 International Journal of Sociology of the Family

states. We use these data sets to examine various theories of domestic violence

cross-nationally.

INDIVIDUAL-AND FAMILY-LEVEL STUDIES

The literature on domestic violence focuses on individuals and families as

units of analysis. A January, 2008, Google Scholar search, using key words "domestic violence," yielded 788,000 articles, most of which focused at these levels. (There may have been exceptions to this rule, but there were none on the first twenty, or last twenty, pages of the list).

Systematic reviews of the literature of family violence in The Journal of Marriage and Family by decade also indicate a focus on individuals and the family. Gelles' (1980) review, in fact, reflected on research of both the sixties and the seventies and stressed that the former tended to see domestic violence

as rare, the result of mental illness and/or poverty, while the latter concentrated

more on measuring the incidence of family violence and factors related to violence in the family. Gelles (1980) also outlined five theories of family violence, one of which, Goode's (1971) Resource Theory (stressing the use of violence by husbands who want to retain a dominant position when they lack other resources, like education and job prestige), was especially useful for accounting for wife abuse at the family level. Gelles and Conte's (1990) review of the eighties literature focused more attention on the sexual abuse of children than the earlier review, but also devoted sections to research that had examined

the changing rates of family violence more generally, the effects of violence on children and women and the assessment of intervention strategies, again at the family level.

Johnson and Ferraro's (2000) review of the nineties literature also focused

on work that dealt with domestic violence at the individual and family levels, though not to the complete exclusion of more macro considerations. Johnson and Ferraro featured a typology of partner violence, distinguishing "intimate terrorism," violence moved by a wish to exert control over a partner, from "common couple violence" (in which one or both partner lash(es) out at the other), "violent resistance" (in which one partner, usually the woman, fights back against an abusing partner), and "mutual violent control" (in which both husband and wife are controlling and violent). Johnson and Ferraro also suggested that a promising development has been work around violence as a mode of control, not only in heterosexual relationships, but also in same-gender relationships.

Unlike previous decade reviews, however, Johnson and Ferrarro did devote

a small fraction of their review to research that looked at global variation in partner violence, including some attention to work suggesting the considerable

variation in the incidence of partner violence by nation (Heise, 1994) and

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms

Determinants of Domestic Violence: A Cross-National Study 149

explanations for some of this variation (see, for instance, McWilliams [1998] work on Northern Ireland and, more generally, "societies under stress.") We will be returning to some of this literature as we develop hypotheses about why this international variation exists.

Research since the last decade review has retained an emphasis on individual-and family-level research, while continuing to bring insights from other societies and hints at what more macro-level studies could look for.

Kacen's (2006) study of Ethiopian immigrants to Israel suggests, among other things, that patriarchal family structures vary in their intensity from one society

to another and for one sub-society over time. Still, the most serious treatment of cross-cultural variation in domestic violence remains that done by David Levinson (1989).

SOCIETAL LEVEL STUDIES OF DOMESTIC VIOLENCE

Levinson in his (1989) Family Violence in Cross-Cultural Perspective, examined domestic violence, using data from The Human Relations Area File. We use Levinson extensively as a source of theories about domestic violence at a societal level. Levinson presents four theories, a Resource Theory, an Exchange Theory, a Culture of Violence Theory, and a Patriarchal Theory, that we believe are testable in a cross-national context and, in fact, will try to test with cross-national data. We will add two more theories of our own, a

Modernization Theory and a Dependency Theory, and will test these two as well. We focus, in our theorizing, on violence perpetrated by husbands upon wives.

RESOURCE THEORY

Resource theory was first espoused by Goode (in 1971) and suggests that the more resources a husband brings to a relationship, the more power he has, but

the less likely he will actually resort to violence. When, however, a map's superior power is threatened by a wife's access to educational or job-related resources, he may resort to violence to re-establish himself as dominant. The implicit view of domestic violence here is consistent with Johnson and Ferraro's

"intimate terrorism" type of violence, where violence is used to exert consistent

control over a spouse, but the latter is clearly focused at the family-level of analysis.

Levinson uses this approach to understand macro level variation in domestic violence. He argues that resource theory implies that societies where men's traditional (relative) power is eroded because of women's increased access to societal resources will involve large-scale status inconsistencies for men. This broad-based experience of status inconsistency, in turn, will be associated with increases in domestic violence generally, Levin posits.

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms

150 International Journal of Sociology of the Family

While Whaley (2001) does not focus on domestic violence directly, her consideration of the relationship between gender inequality, more generally, and rape suggests reason for doubting a direct and consistent relationship between increasing gender equality and violence. In the short term, it makes sense to Whaley that gender equality might pose a threat to the status quo and result in increased rape rates. But in the long term, she argues, gender equality should improve the social "climate" towards women and reduce rape rates. She tests these hypotheses with panel data from 109 U.S. cities, and does, in fact, find support for a negative long-term relationship between gender equality in income, education, access to high-status occupations and legal status, on the one hand, and rape rates. In any case, Whaley's work suggests the possibility

that in cross-sectional analyses, where the cumulative effect of increased gender

equality over time is more likely to show up than shorter-term effects, one may expect to find that increased access to resources will be associated with diminished levels of domestic violence.

Levinson's and Whaley's versions of Resource Theory, then, suggest two, competing hypotheses:

Levinson-based Resource Theory hypotheses

- As levels of education, labor force participation and access to political roles for women increase, domestic violence within that nation will increase.

Whaley-based Resource Theory hypothesis:

- As levels of education, labor force participation and political roles for women increase, domestic violence within that nation will decrease.

EXCHANGE THEORY

Exchange Theory suggests that domestic violence will be particularly high in societies where its benefits to perpetrators are high and particularly low in societies where the costs to perpetrators are low. In many societies costs of violence are low because of inadequate social controls placed on such violence and because an emphasis on male aggressiveness actually encourages it (Levinson 1989; see also Kacen, 2006, for a more recent version). Levinson posits that as the costs of domestic violence rise in a society, the practice of domestic violence will decline. Levinson's view here grows out of Gelles' (1983) individual-level theory that people hit and abuse other family members because they can (Levinson 1989:15). At the national level, we deduce the following:

- Nations that have laws against domestic violence will have lower levels of domestic violence than nations without such laws.

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms

Determinants of Domestic Violence: A Cross-National Study 151

- Nations that do have domestic violence laws, but do not enforce them, will have higher rates of domestic violence than nations that have them and do enforce them.

CULTURE OF VIOLENCE THEORY

Levinson also examines a Culture of Violence theory, first hypothesized by Wolfgang and Ferracuti in 1967 and, as we observed above, revisited more recently by McWilliams (1998) as a "societies under stress" thesis. This theory suggests that violent societies are more likely than nonviolent societies to permit domestic violence partly because where violence is used for conflict resolution generally, it is likely to be accepted as a means of conflict resolution within

the household as well. McWilliams' (1998) version stresses that general violence blocks public awareness about issues of domestic violence. From this perspective we derive the hypothesis that:

- The more recently a nation has been involved in war or internal strife, the higher the levels of domestic violence.

THE PATRIARCHAL THEORY

Levinson, borrowing from Martin (1973) and Dobash and Dobash (1979), also presents a Patriarchal Theory on family violence. This theory submits that, throughout history, males have dominated society and women were to be treated as men's possessions. Patriarchal norms protect men's ability to control their wives and justify their use of violence to do so. These norms have historical roots that emphasize female subordination. There is variation, however, in the extent to which patriarchal norms remain intact. One should be able to find evidence of their continuing presence in laws and practices that restrict women's

freedoms in society at large or within the family particularly. We submit, for instance, that two indicators that a society remains patriarchal are that women's

reproductive rights are inhibited and that great age differences continue exist between husbands and wives. We hypothesize, then, that:

- Countries where women enjoy greater reproductive rights will have lower rates of domestic violence than other countries.

- Countries in which the age difference between men and women at marriage is particularly great will have particularly high levels of domestic violence.

MODERNIZATION THEORY

We argue that modernization should enhance women's status and may be a reason behind a decreasing resource difference between men and women in a society (Resource Theory), increasing costs of doing domestic violence (Exchange Theory), as well as an undermining of traditional patriarchal norms

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms

152 International Journal of Sociology of the Family

(Patriarchal Theory). As Parsons (1960) and Smelser (1966) pointed out, modernization frequently leads to a valuing of universalistic over particularistic norms and achievement over ascription. Hence we expect women (and men) to be freed from traditional gender norms as a country modernizes. To the extent that the freedom is associated with a reduction of

resource differences between men and women, we would expect that, in the long term (Whaley's version of Resource Theory), there would be a diminution in domestic violence. To the extent that societal norms and laws are changed to reflect universalistic and achievement norms and, consequently, to provide penalties for domestic violence (Gelles, 1983, and Levinson, 1989) and to undermine traditional patriarchal norms (Martin, 1973, and Dobash and Dobash, 1979), we would also expect such a diminution. Modernization has been measured (e.g. Clark, 1990) with simple indicators, like gross domestic product per capita and level of urbanization. Consequently, we would expect that:

- There will be less domestic violence in countries with greater gross domestic product per capita.

- There will be less domestic violence in countries with greater urbanization.

ECONOMIC DEPENDENCY THEORY

We also argue that economic dependency, other things being equal, should diminish women's relative status and consequently their access to resources and to favorable norms and laws. Economic dependency has frequently been associated with decreases in women's access to educational, economic and political resources (see Bollen, 1983, Chase-Dunn, 1975, Clark, 1986, Clark, 1991, Clark, Ramsbey and Adler, 1991, Ward 1984, Semyonov and Shenhav, 1988). Dependency theorists do not deny the presence of patriarchal relations in societies before their incorporation into the capitalist world system. But they do suggest that patriarchy is often augmented once capitalism penetrates new areas. Patriarchy is a mechanism through which a balance is maintained between women's roles as producers and reproducers for capitalist markets in goods and labor. An interest in maintaining this balance exists for capitalist enterprise everywhere, but it is in nations that are dependent upon multinational corporate investment that it is most likely to overwhelm countervailing interests.

It is these nations in which governments, for instance, will make the greatest concessions to corporate requirements for profitable environments, limiting, for instance, unwanted regulations on gender discrimination. Consequently, we hypothesize:

- The greater the economic dependency of a country, the higher the overall levels of domestic violence within that country.

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms

Determinants of Domestic Violence: A Cross-National Study 153

MEASUREMENT STRATEGIES

Measuring Domestic Violence Level

Measuring the level of domestic violence in any particular country is tricky. Gelles (1980), for instance, shaped his review of the 1970s literature on family violence around innovations in measurement that had permitted the first decent indications of the incidence in family violence in the United States. One of the

problems with doing cross-national research is the inability to be certain that studies of the incidence of domestic violence produce comparable indicators from country to country. We believe that, despite serious data reliability questions about both of the measures we use here, it is better to get some insight, however questionable, than none at all.

From the State Department's (2005) website we coded descriptions of domestic violence levels by the Department's 2005 Country Reports on Human Rights Practices for 158 nations. Countries were coded 1, 2 and 3 for low, medium and high levels of domestic violence, respectively. We looked for particular descriptors in the qualitative descriptions of each country to determine

our codings. For example, countries that had descriptions about the domestic violence that included such words as domestic violence was "rare" or "had

few cases reported" were coded as a 1 for low level of domestic violence. Thus the State Department's (2005) description for San Marino reported the following regarding domestic violence: "Violence against women, including spousal abuse, was rare" and so we coded this country with a 1. Countries whose description contained words that assessed domestic violence as a "problem" or "significant" were coded with a 2. For example, the for Finland contained the following description: ". . .violence against women, including spousal abuse, continued to be a problem." For country descriptions that contained words such as "widespread" or "serious problem" to describe domestic violence, we used the code of 3. Thus, the report for Albania, for instance, contained the following: "domestic violence against women, including spousal abuse, remained a serious problem." To check the reliability of our interpretations of the State Department's (2005) assessment of domestic violence within countries we used an inter-observer check method. The two

authors read descriptions of 25 of the same countries and our coding for those countries were in agreement for 23 (or 93% of the) cases. Each of us then coded about half of the remaining nations separately. In the end, we decided to combine countries that were coded as having low levels of violence with those that we coded as having medium levels because there were only 10 of the former. We finally ended with 57 countries coded as a 1 for low levels of domestic violence and 111 countries with a 2 for high levels of domestic violence. We call the resulting dichotomous variable Domestic Violence I.

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms

154 , International Journal of Sociology of the Family

Unfortunately, the State Department's assessments of national levels of domestic violence are often impressionistic, perhaps based upon commonplace evaluations within the nation, perhaps, occasionally, based upon population based measurements. They are, as the examples in this paragraph suggest, often explicitly about violence against women, and not about other kinds of domestic violence, but sometimes they are not explicitly confined to statements

about violence against women. They seem to be written by an author who is trying to make national evaluations that are comparable from one country to another, but they are undoubtedly based upon assessments made at the country level, assessments that may or may not be informed by knowledge of domestic violence rates in other countries.

As a second measure of domestic violence, which we call Domestic

Violence II, we use population-based indicators for 28 nations of the percent of women physically assaulted by an intimate partner as reported Heise, Ellsberg

and Gottemoeller (1999). The coverage of the surveys, completed between 1985 and 1999, on which these population-based measures are based is sometimes, but not always, the entire country. When it is clearly not based on the entire country, we make the simplifying assumption that a local rate is a reasonable approximation of a national one. In some cases the rates are based upon a question asking whether a woman has ever been physically assaulted; in some cases, they are based upon a question asking whether she's experienced such an assault in her current relationship. Again, our simplifying assumption has been that such rates are comparable.

Measuring for Women s Resources

We measured women's resources with four indicators: women's labor force

participation rates, the percentage of eligible females enrolled in secondary schools, the percentage of women in the lower house, and the year that women got the vote. The 2003 measure of women's labor force participation rates and the percentage of eligible females enrolled in secondary schools were from the United Nations Development Program's (2004) Human Development Report. A 2001 measure of the percentage of seats women have in the lower house (or only house) of the national legislature is from United Nations Development Program, 2002. The number of years since women received the right to vote is from Inter-Parliamentary Union, 2004.

Measuring Presence of Domestic Violence Laws and whether a Country Enforces Such Laws, if Present

To measure societal disincentives for domestic violence (and therefore "costs"

associated with its commission [Exchange Theory]), we have used the presence of laws against domestic violence and the enforcement of the laws (if present).

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms

Determinants of Domestic Violence: A Cross-National Study 155

We again used reports from the State Department's (2005) 2005 Country Reports on Human Rights Practices: countries without laws against domestic violence were coded 0, while countries with such laws were coded 1. For

example the description for Australia stated "the law prohibits violence against women, including spousal rape and abuse." If the report of a particular country did not refer to laws in any way or specifically said it did not have laws against domestic violence, these countries were coded with a 0 for not having domestic

violence laws. For example, the report on Thailand contained the following" "domestic violence against women was a significant problem, and there were no specific laws addressing the problem." We used a similar procedure for coding whether domestic violence laws, if they existed, were enforced. Again, most of the time the State Department directly stated whether a country did or

did not enforce (or prosecute) the laws for domestic violence. For Latvia, for instance, the State Department's report claimed that "although domestic violence was a significant problem, the government did not effectively enforce the law" and so this country was coded with a 0 for failure to enforce domestic violence laws. Countries that clearly did enforce laws were coded 1 for doing so. We did an inter-coder reliability check whether countries had domestic violence laws and whether laws, when present, were enforced with a sub sample of 25 countries. The two authors agreed 84 percent of the time on the former and 80 percent of the time on the latter.

Measuring Patriarchy

To test Patriarchal Theory we used two measures of the degree of Patriarchy in a society: one is a measure of women's reproductive rights and the other is a measure of men's power within marriage. The indicators we used to create the measure for women's reproductive rights are the same employed by Clark (2006), an indicator produced through factor analysis of an abortion policy index, a maternity leave index, an emergency contraception index and a measure of the prevalence of contraceptive use.1

Another indicator we used to measure how patriarchal a country was the mean age difference between men's mean age at marriage and women's mean age at marriage. On average we expect men to be more powerful in societies where their average age of marriage greatly exceeds that of their wives. The data on both men's mean age at marriage and women's are from the U.N. Statistical Division, 2000.

Measuring the Presence of War or Internal Conflict between 1995 and 2005

We used one measure of whether a nation had a Culture of Violence: whether

it had hosted a war or warlike activity within its borders between 1995 and

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms

156 International Journal of Sociology of the Family

2005. Data were derived from a content analysis of the history provided for each country in the 2008 The World Almanac and Book of Facts (2008). Countries whose almanac description failed to mention war or warlike activity in the period were coded 0; those whose descriptions mentioned war or warlike activity in the period were coded 1,2

Measuring Modernization

We used two independent variables as measurements of how modern a country is: (1) percentage living in urban areas and (2) GNP per capita. Both of these indicators are available for every country in the world in the World Almanac and Book of Facts 2003.

Measuring Economic Dependency

To measure the economic dependency of a nation we used an export diversification index. The export diversification index is from UNCTAD (2001) and is the number of different exported products worth over $100,000 (or greater than 0.3% total national exports) in a country. The index is derived from the "absolute deviation of the country from the world structure" (UNCTAD 2001). A higher score reveals highly concentrated export structures; thus the higher the score the higher the level of economic dependency on other countries.

FINDINGS

Our dichotomous measure of domestic violence (categories = low and high), Domestic Violence I, and the continuous one, based on population surveys, Domestic Violence II, have very different levels of coverage. We have measures

of Domestic Violence I for 156 nations but of Domestic Violence II for only 28 nations. There were only 26 nations on which we had measures for both indicators. Of these 26, only five were classified as low on Domestic Violence I and, among these five, the mean percentage of women reporting having experienced physical assault in population-based surveys was 19.8 percent. Of the 26, 21 were classified as high on Domestic Violence II and, among these 21, the mean percentage of women reporting having experienced physical assault was 34.4 percent. The difference between these two groups is significant at the .02 level, indicating a degree of inter-measure reliability.

We examined the association between both Domestic Violence I and

Domestic Violence II and independent variables meant to represent important dimensions of theoretically meaningful concepts. Table I, for instance, presents the means of several independent variables for countries with low and high levels of domestic violence as captured in Domestic Violence I.

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms

Determinants of Domestic Violence: A Cross-National Study 157

Table I

T-Tests of Domestic Violence I with Various Independent Variables

Independent Low Levels of High Levels of Significance Level Variables Domestic Domestic

Violence I Violence I

Percent of lower house 17.0 13.0 .02

Seats held by Women

Female labor force 52.4 52.6 .97

participation (around 2000)

Percent of eligible females 93.0 62.9 .001

enrolled in secondary school

The year women got the vote 1934.6 1951.4 .001

Mean age difference at 3.2 3.7 .25

marriage of women from men

Women's reproductive rights .48 -0.18 .001

Diversification index 2001 .56 .72 .001

% living in urban areas (2003) 63.4 48.8 .001

GDP per capita (2003) 15450.9 5507.2 .001

You will recall that the Resource Theory led us to contrary expectations about how women's access to important societal resources might be related to domestic violence. Levinson's view was that, as women's access to such resources increased, there would be more domestic violence, while Whaley's view was that in societies where women's access to such resources is high, domestic violence would be low. Table I provides more support for Whaley's formulation of the Resource theory than for Levinson's. For instance, in countries with low levels of domestic violence, about 17.0% of parliamentary seats were held by women, while in countries with high levels of domestic violence, the average was 13.0% of seats. Thus, countries where women occupied more seats in parliament were more likely to have low levels of domestic violence than countries where fewer seats were held by women. This association is significant at the .02 level. Two other indicators of women's resources, their access to secondary education and the number of years they've had the vote, are also negatively, and significantly, associated with their experience of domestic abuse, as measured by Domestic Violence I. Only one indicator of their resources, their access to the formal labor force, is not significantly associated with Domestic Violence I.

The Patriarchal Theory led us to expect there would be higher levels of domestic violence in societies that perpetuate the subordination of women. We used two independent variables as indicators of this subordination, one being women's access to reproductive rights and the other being the mean age

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms

158 International Journal of Sociology of the Family

difference of men and women at the time of marriage. Table I shows that both of these measures are related to Domestic Violence I in the direction anticipated

by Patriarchal Theory, but that only the index of women's reproductive rights is significantly associated with it.

From Depehde^cy Theory, we derived the suggestion that higher values of the diversification index (indicating lower levels of product diversity) should

be associated with high levels of Domestic Violence. Table I suggests that such is the case.

Modernization Theory led us to expect that the percentage living in urban

areas (2003) and the GDP per capita (2003) would be associated with low levels of domestic violence. Both expectations were borne out, as shown in Table I.

Exchange Theory led us to expect that the presence of laws against domestic violence would be associated with lower levels of domestic violence and that,

if countries had such laws, ones with high levels of enforcement would have lower levels domestic violence than ones with low levels of enforcement. Table

II shows that, in fact, countries with laws against domestic violence were much

less likely to have high levels of Domestic Violence I than countries without such laws and that this relationship is significant.

Table II

Association of Presence of Domestic Violence Laws and Level of

Domestic Violence I

Country has laws against domestic violence

Low level of Domestic Violence I

High Level of Domestic Violence I

No 12(18.5%) 53(81.5%) Yes 43(44.3%) 54(55.7%) Cramer's V .27

Significance Level .001

Moreover, Table III shows that countries that enforce domestic violence

laws are much less likely to evince high levels of Domestic Violence I than countries that do not enforce such laws.

Table III

Association of Enforcement of Domestic Laws and Level of

Domestic Violence I

Country enforces domestic violence laws

Low Level of Domestic Violence I

High Level of Domestic Violence I

No, 18(32.1%) 38(67.9%) Yes 24(63.2%) 14(36.8%) Cramer's V .31

Significance Level .001

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms

Determinants of Domestic Violence: A Cross-National Study 159

The Culture of Violence Theory suggests that domestic violence levels will be higher in societies that have been recently involved in wars or serious outbreaks of internal violence. Table IV presents evidence in support of this theory, inasmuch as it shows that countries that were involved in recent wars or internal strife were significantly more likely to score high levels of domestic Violence I than other countries.

Table IV

The Association of General Violence with Domestic Violence I

Country recently Low Level of Domestic High Level of Domestic involved in war or Violence I Violence I

warlike activity

No 50(42.4%) 68(57.6%) Yes 7(14.0%) 43(86.0%) Cramer's V .27

Significance Level .000

Table V presents a logistic regression analysis for 93 cases, as well as correlation and linear regression analysis. It extends the analysis of the association between Domestic Violence I and various independent variables. Logistic regression involves the assumption, in this case, that the dependent variable is dichotomous and enables the assessment of the controlled

associations of each independent variable with the dependent variable. The first column in Table V, for instance, shows that the association between women's political power (a combination of women's representation in parliament and the year they got the vote and meant to measure women's resources) with Domestic Violence I (B = .14) is not significant, controlling for other variables in the equation. Similarly, the presence of laws against domestic violence in a nation (our measure here of costs associated with domestic violence) has a non-significant association with Domestic Violence I, as does our measure of reproductive rights (B =.02), and the presence of a violent subculture, or wars (B = -.42). In fact, the only variables that show a significant association with domestic violence are economic development (B=-1.00), a measure of the modernization of a society (a combination of urbanization level and gross domestic products per capita), and economic diversification (B =4.3), measuring its dependency level. This analysis suggests that modernization and dependency may underlie the empirical connections found earlier to support the four other theories we considered: the Resource, Exchange, Culture of Violence and Patriarchal theories.

The linear regression analysis, shown in the second column of Table V, essentially replicates the logistic regression analysis of the first column, despite

the fact that the dichotomous dependent variable is more suitable for the latter

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms

160 International Journal of Sociology of the Family

Table V

Logistic and Linear Regressions of Domestic Violence I on Independent Variables, as well as Zero-Order Pearson Correlations for the 93 Cases*

Independent Variables Logistic Regression

B

(Standard Error) (Significance level)

Linear Regression

Beta

(Significance Level)

Zero-Order Pearson

Correlation

Pearson r

(Significance)

Women's Political Power .14 -.02 -.28

(.36) (•56) (.006)

Presence of Laws against -.16 .00 -.21 Domestic Violence (.72) (.85) (.039)

Women's Reproductive Rights .02 .01 -.32

(.41) (.92) (.002)

Presence of Wars between -.42 -.06 .15

1995 and 2005 (.74) (.58) (.16) (•57)

Economic Diversification 4.3 .30 .48

(1.8) (.018) (.000) (.02)

Economic Development -1.00 -.35 -.48 (•39) (.012) (.000) (■01)

Nagelkerke r square 0.38

R square .29

N 93 93 93

than the former. Thus encouraged, we examined the zero-order associations between the independent variables for the 93 cases (see column three in Table 5) to compare them with the betas in the linear regression. Notably, the zero-order associations for the 93 cases provide support for all six of the theories that we found support for above, except the Culture of Violence Theory inasmuch as the association between the presence of war and Domestic Violence I is not statistically significant for the 93 cases, even though it remains positive (r =.15). Again what jumps out is that, while all the betas suggest weaker controlled associations than the zero-order ones, only the controlled associations for economic development (beta =-.35) and economic dependency (beta =.30) remain significant and strong, with other independent variables controlled. This comparison suggests, again, that modernization and dependency may be underlying reasons why women's resources, legal penalties against domestic violence, women's reproductive rights and the presence of a violent society are associated with domestic violence: all are affected by

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms

Determinants of Domestic Violence: A Cross-National Study 161

modernization and dependency, while the latter are having independent effects upon domestic violence.

Because there are so many fewer countries for which Domestic Violence II (n = 8) is available than for Domestic Violence I (n = 158), no such multivariate analysis is possible for Domestic Violence II, the population based estimates of the percentage of women who have experienced physical assault. But Table VI shows the zero-order association between Domestic

Violence II and the independent variables, whose association with Domestic Violence I was shown above. These associations are remarkably consistent with those shown for Domestic Violence I.

Table VI

Correlations of Domestic Violence II and Important Independent Variables

Percent of Lower House Parliamentarians

who are Women

Female Labor Force Participation

Percent of Eligible Females Enrolled In Secondary Schools

Year Women Got the Vote

Difference Between Men's Mean

Age at Marriage and Women's

Women's Reproductive Rights

Diversification Index

Gross Domestic Product per capita

Urbanization level

Presence of Laws against Domestic Violence

Enforcement of Anti-Domestic Laws

War during 1995-2005 period

Domestic Violence II

(Number of Case In Parentheses)

-.46*

(28)

-.21

(27)

-.48*

(23)

.33

(21~\

.42

(19) -.38*

(28) .46*

(26)

. 49**

(28)

-.37*

(28) .13

(25)

-.60**

(18)

.29

(28)

Notes: ** indicates significant at .01 level; *, at .05 level.

Thus, Domestic Violence II is significantly and negatively associated with two measures of women's resources (the percent of parliamentarians who are

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms

162 International Journal of Sociology of the Family

women and women's involvement in secondary education). The measure of patriarchy that was significantly associated with Domestic Violence I— women's reproductive rights—is also significantly and negatively associated with Domestic Violence II: countries that have provided women with more legal reproductive protections are also ones in which women report the least physical assault in their marriages. Again, both our measures of economic dependence and economic development were strongly associated with Domestic Violence II, as they were with Domestic Violence I. Presence of laws against domestic violence is associated with lower values of Domestic Violence II, as it was for Domestic Violence I, but the enforcement of laws, where they exist, is strongly (r =-.60) and significantly associated with it. Finally, war in the last ten years was positively associated with higher values of Domestic Violence II, as it had been with Domestic Violence I, but this time it is not significantly associated with it.

In short, the pattern of associations supports all of the theoretical perspectives when Domestic Violence II is used as the dependent variable, as it did when Domestic Violence I was the criterion. Both indicators of domestic

violence have their weaknesses, but both are associated with the same set of

dependent variables and this lends additional credibility to the inferences that societies where women have access to valued resources, where the legal system promises negative consequences for those who commit domestic violence, where patriarchal norms have been eroded, where a culture of violence does not exist, where the economy is not economically dependent and where the economy is reasonably well-developed—where these various conditions exist, domestic violence will be lower than in other societies.

SUMMARY AND DISCUSSION

The available literature on domestic violence led us to numerous studies at the

individual or family level and only a few at the societal level. Individual or family level studies have been used to develop typologies of domestic violence as well as to determine conditions that make some families more likely than others to display domestic violence within any given society. Domestic violence, these studies suggest, is more likely to occur, for instance, in families where male partners abuse drugs and alcohol, are unemployed, or have relatively little education. But we believe that there is variation in domestic violence not only from household to household, family to family, but also from nation to nation and that a complete understanding of the phenomenon of domestic violence cannot be attained without some understanding of this international variation. We have investigated, in this paper, some of the societal characteristics that may determine variation in the levels of domestic violence from one nation to another. We used two measures of domestic violence within

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms

Determinants of Domestic Violence: A Cross-National Study 163

a nation: a) an indicator of whether or not a nation had high levels of domestic violence, derived from a content analysis of the United States State Department's (2005) Country Reports on Human Rights Practices for 158 nations; and b) a measure of the percentage of women who had experienced physical assault at the hands of their partner(s) based on population-based surveys in 28 countries. These two measures are associated with each other, but not so strongly as to suggest they are measuring the exactly the same thing. The first measure, after all, is not in all cases explicitly confined by the

State Department reports to physical assault against wives. Their association does suggest, however, that they tap something similar and we use them, in the absence of more clearly valid and reliable measures, to gain some insight into cross-national variation, where none was previously available.

The two measures yield complementary support for six theories addressing

such variation. We found support for a version of a Resource Theory (Whaley, 2001) that suggested, as women's access to important resources increased, domestic violence would decrease. We also found support for Levinson's (1989) formulation of the Exchange Theory, implying that the presence and enforcement of domestic violence laws in a nation would result in less domestic

violence. Patriarchal Theory led us to the expectation that societies that retain patriarchal norms are more likely to exhibit high levels of domestic violence than societies that have challenged such norms and we found strong and significant support for this view, though more when patriarchal norms were measured in terms of women's access to reproductive rights than in terms of their marital power. A Culture of Violence theory suggested that there would be an association between domestic violence levels and the presence of violence in the society at large and we found support for this view when we measured a culture of violence in terms of whether a country was involved in war or warlike activity in a 10 year span, though such support was forthcoming when we measured domestic violence in terms of Domestic Violence I, not in terms

of Domestic Violence II. Our Economic Dependency Theory led us to expect that domestic violence levels would be higher in countries that were economically dependent than those that were not and we found strong support for this expectation, as we did for ones based on our Modernization Theory.

To assess which of the six theories of cross-national domestic violence

levels had the greatest explanatory value, we did logistic and linear regression analyses, as well as a correlation analysis based upon the same number of countries, involving Domestic Violence I (the indicator for which we had information on enough nations to permit such a test). These analyses suggest that only 2 of the 6 variables have significant associations with domestic violence when the other variables are controlled: measures of modernization

and economic dependency. We infer that modernization and dependency may

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms

164 International Journal of Sociology of the Family

underlie the empirical connections found earlier to support the four other theories we considered: the Resource, Exchange, Culture of Violence and Patriarchal theories. Women's resources, laws against domestic violence and their enforcement, women's reproductive rights and levels of internal violence within a society MAY all be spuriously related to levels of domestic violence.

We speculate that modernization and economic dependence may have opposite effects on women's resources, the legal costs associated with domestic violence, levels of patriarchy and levels of internal violence, as well as levels of domestic violence. Modernization, Parson (1960) and Smelser (1966) posited, leads to a valuing of universalistic over particularistic norms and achievement over ascription and should, we argue, make it possible for women to gain access to valued resources, for laws against women's subjugation to be passed, and even for the diminished use of violence and the increased use of rational-legal means of conflict resolution. To the extent that universalistic norms and achievement orientation are internalized and to the extent that they undermine traditional justifications for domestic violence, we would expect their existence to lead to a diminished use of violence within families as well.

Economic dependency, on the other hand, has been linked to a reliance on traditional patriarchal norms to justify women's roles as producers and reproducers for capitalist markets, and so, we would argue, would limit their access to economic, educational and political resources, as well as to legal protections, and might engender an even greater reliance on violence for conflict resolution, both in and out of the home.

Although we've begun the examination of cross-national variation of domestic violence, our research is plagued by limitations. For instance, both of our measures of domestic violence have weaknesses. Domestic Violence I

is based upon our content analysis of U.S. State Department's (2005) Country reports, most of which are based more on commonplace impression than population-based surveys and many of which may be colored by political agendas of, say, State Department representatives. Domestic Violence II is, in fact, based on population based surveys, but these were done at different times, using different methodologies, with different levels of coverage, and these are therefore of questionable comparability.

Our investigation has brought us to a belief that economic dependence and modernization may be crucial to an understanding of international variation in domestic violence. Future cross-national research should investigate this finding with more complete and reliable data sets, as such data sets become available.

Notes

1. The abortion policy index is based on eight conditions under which abortion is allowed in each country: (1) not legally at all, (2) to save the life of the woman, (3) to preserve physical health, (4) to preserve mental health, (5) result of rape or incest, (6) in the

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms

Determinants of Domestic Violence: A Cross-National Study 165

case of fetal impairment, (7) for economic or social reasons, and (8) abortion requested. Data on abortion policies are from the United Nations Development Program (2000). The indicator of maternity leave used is an additive scale of two variables: (1) the proportion of normal wages one receives during the covered period and (2) the number of days one is entitled to for maternity leave divided by 150. Data on maternity leave policies are from United Nations (2001). The measure of women's reproductive rights is also made up of a nation's emergency contraceptive marketing status. Data on emergency contraception are from the Office of Population Research (2005). A nation's status for emergency contraception is coded at four levels: (1) countries where emergency contraception is not marketed were coded with a 0, (2) countries where emergency contraception is sold by prescription only with a 1, (3) countries where the contraception can be bought directly from a pharmacist with a 2 and (4) in countries where the emergency contraception can be bought over the counter with a 4. The final variable used to make up the measurement of women's reproductive rights is the prevalence of contraceptive use. Data for the prevalence of contraceptive use is from the United Nations Population Fund (2004).

2. One country we coded 0 was Guyana, on the northern coast of South America. Its total modern history, in the Almanac, was:

When Pres. Jagan Mar. 6, 1997, Prime Min. Samuel Hinds succeeded him. Jagan's widow, Janet, became prime min. Mar. 17. She won the presidency in a disputed election Dec. 15. She resigned because of ill health Aug. 11, 1999, and was succeeded by Bharrat Jagdeo, then 35, who became the youngest head of state in the Americas. He won reelection Mar. 9,2001, and Aug. 28, 2006.

Floods from torrential rains, Jan. 2005, affected about 40% of the

population. Gunmen in Georgetown killed Agric. Min. Satyadeow Sawh and 2 members of his family, Apr. 22, 2006.

We did not see the assassination of a government minister in this description as sufficient

evidence of widespread violence to classify Guyana as a nation in the grip of a culture of violence.

Contrast this description with a small portion of the recent history attributed to Liberia,

on the southwest coast of Western Africa, a country that we coded as gripped by violence and therefore coded as a 1:

A civil war began Dec. 1989. In Sept 1990, Pres. Doe was captured and put to death. Despite the introduction of peacekeeping forces from several countries, the conflict intensified. Factional fighting devastated Monrovia in Apr. 1996. On Sept. 3, Ruth Perry became modern Africa's first female head of State, leading a transitional government. By then, the civil war had claimed more than 150,000 lives and uprooted over half of the population.

References

Bollen, Kenneth (1983). World System Position, Dependency, and Democracy: The Cross National Evidence. American Sociological Review 48: 468-479.

Brush, Lisa (1990). Violent Acts and Injurious Outcomes in Married Couples: Methodological issues in the National Survey of Families and Households. Gender & Society 4(1): 56-67.

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms

166 International Journal of Sociology of the Family

Chase-Dunn, Christopher (1975). The Effects of International Economic Dependence on Development and Inequality. American Sociological Review 40: 720-38.

Clark, Roger (1991). Women's Access to Prestigious Occupations: A Cross-National Investigation. Social Science Quarterly 72(1): 20-32.

Clark, Roger, Thomas Ramsbey, and Emily Stier Adler. Culture, Gender, and Labor Force Participation. Gender & Society 5(1): 47-66.

Dobash, Emerson R. and Dobash, Russell P. (1998). Rethinking Violence against Women. Thousand Oaks, CA: Sage Publications

Dobash, Emerson R., and Dobash, Russell P. (1979). Violence against Wives. New York: Free Press.

Galtung, Johann (1971). A Structural Theory of Imperialism. Journal of Peace Research 2: 89-117.

Gelles, Richard J. (1980). Violence in the Family: A Review of Research in the Seventies. Journal of Marriage and the Family. 42: 873-885.

Gelles, Richard J. (1983). An Exchange/Social theory, in D. Finkelhor, R.J. Gelles, G.T. Hotaling, and M.A. Straus (eds.) The Dark Side of Families: Current Family Violence Research. Beverly Hills, CA: Sage, pp. 151-165.

Gelles, Richard J., and Cone, Jon R. (1990). Domestic Violence and Sexual Abuse of Children: A Review of Research in the Eighties. Journal of Marriage and the Family 52: 1045-1058.

Goode, W. (1971). Force and Violence in the Family. Journal of Marriage and the Family 33: 624-636.

Heise, L., Raikes, A., Watts,C. H., and Zwi, A. B. (1994). Violence against Women: A Neglected Public Health Issue in Less Developed Countries. Social Science and Medicine 39: 1165-1179.

Heise, L. Ellsberg, M., and Gottemoeller, M. (1999). Ending Violence Against Women. Population Reports, Series L, No. 11. Baltimore, Johns Hopkins University School of Public Health, Population Information Program, December.

Inter-Parliamentary Union (2004). Women's Suffrage. Retrieved on 4/15/04. http:// www.ipu.org/wmn-e/suffrage.html

Johnson, Michael P., and Ferraro, Kathleen J. (2000). Research on Domestic Violence in the 1990s: Making Distinctions. Journal of Marriage and the Family 62: 948-963.

Kacen, Lea (2006). Spousal Abuse Among Immigrants from Ethiopia in Israel. Journal of Marriage and Family 68: 1276-1290.

Levinson, David (1989). Family Violence in Cross-Cultural Perspective. Newbury Park, CA: Sage Publications.

MacMillan, Ross, and Gartner, Rosemary (1999). When She Brings Home the Bacon: Labor Force Participation and the Risk of Spousal Violence Against Women. Journal of Marriage and the Family 61: 947-958.

Martin, D. (1983). Battered Wives. New York: Pocket Books.

McWilliams, Monica (1998). Violence against Women in Societies Under Stress, in R. Emerson Dobash and Russell P. Dobash. Rethinking Violence against Women. Thousand Oaks, CA: Sage, pp. 111-140.

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms

Determinants of Domestic Violence: A Cross-National Study 167

Office of Population Research, Princeton University (2005). Dedicated ECPs Worldwide. List maintained by James Trussel and Lisa Wynn and supplied by Lisa Wynn ([email protected]), March 2005.

Parsons, Talcott. Structure and Process in Modern Societies (1960). Chicago: Free Press.

Semyonov, Moshe, and Y. Shenhav (1988). Investment Dependence, Economic Development, and Female Opportunities in Less Developed Countries. Social Science Quarterly 69: 961-78.

Smelser, Neil (1966). The Modernization of Social Relations, in M. Weiner, ed., Modernization. New York: Basic Books, 110-121.

United States State Department (2005). 2005 Country Reports on Human Rights Practices. Public Communication Division, U.S. Department of State. Retrieved 6/07. http:// www.contact-us.state.gov

United Nations (2001). The World's Women 2000: Trends and Statistics. New York: United Nations.

United Nations Conference in Trade and Development (UNCTAD). 2001. UNCTAD Handbook of Statistics On-line. Retrieved 3/15/05. www.unctad.org/Templates/ Page, asp ?intItemID=1890

United Nations Development Programme. 2000. World Abortion Policies 1999. Retrieved 3/15/05. http://www. un. org/esa/population/publications/abt/fabt. html.

United Nations Development Program (2004). Human Development Report 2004: Deepening Democracy in a Fragmented World. New York: Oxford University Press. Retrieved 3//05. www.un.org/hdr2004/www.un.org/hdr2002/.

United Nations Population Fund (2004). State of World Report 2004. Retrieved 4/1/2005. http://www. unfpa. org/swp/2004/pdf/en_swp04.pdf

United Nations Statistics Division (2000). The World's Women 2000: Trends and Statistics. Table 2.A: Indicators on Marriage and Marital Status. Retrieved 5/25/2004. http:// unstats. un. org/unsd/demographic/wws000/table2a. html

Whaley, Rachel B. (2001). The Paradoxical Relationship Between Gender Inequality and Rape. Toward a Refined Theory. Gender & Society 15 (4): 531-555.

Wolfgang, M. E., and F. Ferracuti (1967). The Subculture of Violence: Toward an Integrated Theory of Criminology. London: Tavistock.

The World Almanac and Book of Facts (2008). New York: World Almanac Books.

World Almanac Book of Facts (2003). New York: World Almanac Education Group, Inc.

This content downloaded from 216.145.162.122 on Thu, 25 Apr 2019 20:07:49 UTC All use subject to https://about.jstor.org/terms