Citation
Examining Changes in Male and Female Intimate Partner Homicide over Time, 1990-2000

Material Information

Title:
Examining Changes in Male and Female Intimate Partner Homicide over Time, 1990-2000
Creator:
Reckdenwald, Amy
Place of Publication:
[Gainesville, Fla.]
Florida
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (129 p.)

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Criminology, Law, and Society
Criminology, Law and Society
Committee Chair:
Lanza-Kaduce, Lonn M.
Committee Co-Chair:
Parker, Karen F.
Committee Members:
Piquero, Alexis R.
Zsembik, Barbara A.
Graduation Date:
8/9/2008

Subjects

Subjects / Keywords:
Domestic violence ( jstor )
Homicide ( jstor )
Intimate partners ( jstor )
Marginalization ( jstor )
Poverty ( jstor )
Shelters ( jstor )
Single status ( jstor )
Violence ( jstor )
Women ( jstor )
Womens studies ( jstor )
Criminology, Law and Society -- Dissertations, Academic -- UF
Genre:
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Criminology, Law, and Society thesis, Ph.D.

Notes

Abstract:
Research on intimate partner homicide has increased in the recent years. This is partially due to the dramatic decline witnessed over the last couple of decades in these types of homicides as well as the growth that has occurred in public awareness and policy responses toward domestic violence. Recent intimate partner homicide research has predominately focused around two perspectives to explain the relationship between intimate partner homicide and domestic violence resources ? the exposure reduction hypothesis and the backlash or retaliation hypothesis, with results that support both (Dugan, Nagin, and Rosenfeld, 1999; 2003). The exposure reduction hypothesis proposes that domestic violence resources that reduce the exposure or contact between intimate partners should decrease the probability of intimate partner homicide, while the backlash hypothesis suggests the opposite. That is, domestic violence interventions may have unintended consequences and increase the risk of intimate partner homicide if they threaten male dominance and control over their partners. The contradictory results from these studies have frustrated advocates and have made them question their efforts to make females safer. Societal remedies to lower intimate partner homicide seem to be addressing male-perpetrated intimate partner violence and homicide insufficiently. It is important to gain a better understanding of what factors are truly influencing gender-specific intimate partner homicide, during a time that is marked by significant transformations in domestic violence legislation and response toward domestic violence with the enactment of the Violence Against Women?s Act of 1994. This research helps to address the contradictory findings while controlling for a number of structural factors that have shown to be important in the homicide literature. The current study examines both the arguments of exposure reduction and backlash, in addition to economic deprivation and marginalization to explain the observed patterns in male- and female-perpetrated intimate partner homicide over time, something that has not been done in the literature to date. Measures of key concepts will be collected for 2 decennial points (1990 and 2000). Supplemental homicide files for this time period and census data will be utilized. Poisson regression models will be used to investigate which theoretical perspective is associated with male- and female-perpetrated intimate partner homicide in 1990 and 2000. Pooled cross-sectional fixed effect time series regression will be used to determine whether changes from 1990 to 2000 in key structural indicators influence trends in male- and female-perpetrated intimate partner homicide during this same time period. Overall, results suggest that key theoretical indicators do in fact influence the trends in female-perpetrated intimate partner homicide, but not male-perpetrated intimate partner homicide. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (Ph.D.)--University of Florida, 2008.
Local:
Adviser: Lanza-Kaduce, Lonn M.
Local:
Co-adviser: Parker, Karen F.
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2010-08-31
Statement of Responsibility:
by Amy Reckdenwald.

Record Information

Source Institution:
UFRGP
Rights Management:
Copyright Reckdenwald, Amy. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Embargo Date:
8/31/2010
Classification:
LD1780 2008 ( lcc )

Downloads

This item has the following downloads:


Full Text





Block, Carolyn R. 2000. The Chicago Women's Health Risk Study. Risk of Serious Injury or
Death in Intimate Violence: A Collaborative Research Project. Final Report to the
National Institute of Justice. Available at
http://www.icjia.state.il.us/public/pdf/cwhrs/cwhrs.pdf.

Boritch, Helen, and John Hagan. 1990. A century of crime in Toronto: Gender, class, and
patterns of social control, 1859 to 1955. Criminology 28:567-600.

Box, Steven, and Chris Hale. 1983. Liberation and female criminality in England and Wales.
British Journal of Criminology 23:35-49.

Box, Steven, and Chris Hale. 1984. Liberation/Emancipation, economic marginalization, or less
chivalry: The relevance of three theoretical arguments to female crime patterns in
England and Wales, 1951-1980. Criminology 22:473-497.

Braithwaite, John. 1979. Inequality, Crime, and Public Policy. London: Routledge & Kegan
Paul.

Browne, Angela. 1985. Assault and homicide at home: When battered women kill. In Advances
in Applied Social Psychology, vol. 3, eds. Michael J. Saks and Leonard Saxe.
Hillsdale, NJ: Lawrence Erlbaum.

Browne Angela. 1986. Assault and homicide at home: When battered women kill. In Advances
in Applied Social Psychology, eds. Michael J. Saks and Leonard Saxe. Hillsdale, NJ:
Erlbaum.

Browne, Angela. 1987. When Battered Women Kill. New York: Free Press.

Browne, Angela, and Robert L. Flewelling. 1986. Women as victims or perpetrators of homicide.
Presented at the Annual Meeting of the American Society of Criminology, Atlanta,
October 29-November 1.

Browne, Angela, and Kirk R. Williams. 1989. Exploring the effect of resource availability and
the likelihood of female-perpetrated homicides. Law and Society Review 23:75-94.

Browne, Angela, and Kirk R. Williams. 1993. Gender, intimacy, and lethal violence. Gender &
Society 7:78-98.

Browne, Angela, Kirk R. Williams, and Donald G. Dutton. 1999. Homicide between intimate
partners: A 20-year review. In Homicide: A Sourcebook of Social Research, eds. M.
Dwayne Smith and Margaret A. Zahn. Thousand Oaks, CA: Sage.

Brush, Lisa D. 1990. Violent acts and injurious outcomes in married couples: Methodological
issues in the National Survey of Families and Households. Gender & Society 4:56-67.

Byrne, James M., and Robert J. Sampson. 1986. The Social Ecology of Crime. New York:
Springer-Verlag.









argued that domestic violence against women was not seen as a national social problem until the

enactment of the 1994 VAWA. The VAWA significantly increased funding for domestic

violence services, "and supported domestic violence specialization in local police departments

and prosecutor offices" (Dugan et al., 2003: 171). Some examples of the VAWA's programs

include policies to encourage the prosecution of abusers, victim services, and domestic violence

prevention programs. Although this act specifically acknowledges violence against women,

violence, both lethal and non-lethal, should be acknowledged against both men and women.

Domestic violence services should influence intimate partner homicide by limiting the

opportunity for contact between violent intimate partners and thus limit the opportunities for

partners to kill each other.

Despite the reports that the increased availability of domestic violence services and

resources is a significant factor in the explanation of the decline in intimate partner homicide

over the past couple of decades (Browne and Williams, 1989; Dugan et al., 1999; 2003),

resources have historically been directed at females and more so married females (Browne and

Williams, 1993; Browne, Williams, and Dutton, 1999). This gives females options besides

killing their partners and may in fact be a main reason that male intimate partner homicide

victimization has decreased significantly. However, this explanation does not hold for female

intimate partner homicide victimization. That is, female intimate partner homicide victimization

has not seen as drastic of a reduction as male intimate partner homicide victimization.

Taken together, there have been many changes witnessed over the last couple of decades

in intimate partner homicide as well as in domesticity, the status of women, economic

deprivation, and in the availability of domestic violence resources. It is important to determine if

these changes can account for the significant decline in intimate partner homicide. Examination









CHAPTER 5
DISCUSSION AND CONCLUSIONS

The goal of this study was to examine the trends in both male- and female-perpetrated

intimate partner homicides from 1990 to 2000 to obtain a better understanding of the differential

declines in male- and female-perpetrated intimate partner homicide. Clearly, the decline in intimate

partner homicide over time is a very complex issue. This research suggests that one explanation

will not suffice. This research was based on the theories of exposure reduction, backlash, and

economic deprivation and marginalization. Results suggest a lack of support for these theories.

Cross-sectional analyses are presented to evaluate the ability of the exposure reduction perspective,

the backlash perspective, and ideas behind economic deprivation and marginalization in explaining

male- and female-perpetrated intimate partner homicide counts in large cities at two time periods

(1990 and 2000). Pooled cross-sectional time series analyses are presented to evaluate the

effectiveness of the exposure reduction perspective, the backlash perspective, and economic

deprivation and marginalization in explaining the changes in male- and female-perpetrated intimate

partner homicide in large cities over time.

Discussion

Dramatic changes were occurring in domesticity, the status of females in terms of income,

employment and educational attainment, economic deprivation, and domestic violence resources

from 1990 to 2000 in large cities. For instance, during this time period the percentage of the

population divorced decreased for both males and females, while the percentage of unmarried

households increased. Furthermore, females have made gains in educational attainment and

employment relative to males (i.e., status of females increased). However, this is not true for

females' median income. Females' median income relative to males actually decreased during this

time period (i.e., status of females decreased). Also, economic deprivation in terms of poverty,









DeWees Mari A., and Karen F. Parker. 2003. Women, region, and types of homicide. Homicide
Studies 7:368-393.

Dobash, Russell P., R. Emerson Dobash, Margo Wilson, and Martin Daly. 1992. The myth of
sexual symmetry in marital violence. SocialProblems 39:401-421.

Dugan, Laura, Daniel S. Nagin, and Richard Rosenfeld. 1999. Explaining the decline in intimate
partner homicide: The effects of changing domesticity, women's status, and domestic
violence resources. Homicide Studies 3:187-214.

Dugan, Laura, Daniel S. Nagin, and Richard Rosenfeld. 2003. Exposure reduction or retaliation?
The effects of domestic violence resources on intimate-partner homicide. Law &
Society Review 37:169-198.

Durose, Matthew R., Carolyn Wolf Harlow, Patrick Langan, Mark Motivans, Ramona R.
Rantana, and Erica L. Smith. 2005. Family Violence Statistics. Washington D.C.:
Department of Justice, Bureau of Justice Statistics.

Ellis, Lee, and Charles Beattie. 1983. The feminist explanation for rape: An empirical test.
Journal of Sex Research 19:74-93.

Fagan, Jeffrey, and Angela Browne. 1994. Violence between spouses and intimates: Physical
aggression between women and men in intimate relationships. In Understanding and
Preventing Violence. Volume 3: Social influences, eds. Albert J. Reiss, Jr., & Jeffrey
A. Roth. Washington, DC: National Academy Press.

Felson, Richard B. 2006. Is violence against women about women or about violence? Contexts
5:21-25.

Flewelling, Robert L., and Kirk R. Williams. 1999. Categorizing homicides: The use of
disaggregated data in homicide research. In Homicide: A Sourcebook of Social
Research, eds. M. Dwayne Smith and Margaret A. Zahn. Thousand Oaks, CA: Sage.

Fox, James Allan. 2005. Uniform Crime Reports [United States]: Supplementary Homicide
Reports, 1976-2002 [Computer file]. Compiled by Northeastern University, College of
Criminal Justice. ICPSR ed. Ann Arbor, MI: Inter-university Consortium for Political
and Social Research [producer and distributor], 2005.

Fox, James Allan. 2004. Missing data problems in the SHR. Homicide Studies 8:214-254.

Fox, James Alan, and Marianne W. Zawitz. 2006. Homicide Trends in the United States.
Washington, DC: Bureau of Justice Statistics. Available at www.ojp.usdoj.gov/bjs/

Gallup-Black, Adria. 2005. Twenty years of rural and urban trends in family and intimate partner
homicide. Homicide Studies 9:149-173.









other options besides resorting to killing their male partners. Also, increased status would reduce

females' dependence on males. Dugan et al. (1999) did indeed find that females' improved status

was associated with intimate partner homicide victimization, particularly male intimate partner

homicide victimization. That is, the increase in females' relative income is associated with a

decline in married female-perpetrated homicide. Furthermore, an increase in females' relative

educational attainment is associated with a decline in non-married male victimization. They

suggest that "more educated women are better able, and perhaps more willing, to exit violent

relationships and thus avoid killing their partner" (204-205).

Exposure Reduction Predictions

In summary, it appears that women in violent relationships may be looking for other

options, but without other options than remaining in the relationship available may turn to lethal

methods. Ideas behind the exposure reduction perspective suggest that limiting the exposure or

contact of intimate partners to one another should decrease the probability of intimate partner

homicide, because there is less exposure to a violent partner and more opportunities to exit the

relationship. Empirically, if exposure reduction efforts were available to intimate partners before

either partner took the violence to a lethal level intimate partner homicide should decrease.

Based on these ideas it seems likely that factors that reduce contact or exposure between intimate

partners in violent relationships should have a significant impact on female-perpetrated intimate

partner homicide, because certain factors may give women other options than having to resort to

violence, lethal or not, to protect themselves.

The availability of domestic violence resources and the possible effect on intimate partner

homicide has gained increase attention. Results have been supportive of the exposure reduction

perspective; however this is only true of female-perpetrated intimate partner homicide. It appears

that domestic violence services that are designed to make females safer are actually making









as sharply as female-perpetrated intimate partner homicide over time despite the increase in

domestic violence resources geared towards preventing violence against females. It is essential to

determine what really has influenced male-perpetrated intimate partner homicide over time. This

may be the only way to for us to see a larger decline in these types of homicides.









increase the female intimate partner homicide victimization rate. Dugan et al. (1999) suggests

that some men are threatened by their loss of power over their female partners as their status

increases.

The backlash perspective theorizes that increased male violence towards their intimate

partners is due to a perceived loss of power or control because of women's improving economic

conditions (Browne, 1987; Vieraitis and Williams, 2002). This hypothesis was first developed by

Williams and Holmes (1981) and Russell (1975) as a warning of the potential consequences of

women's increased economic conditions or gender equality during the women's movement. That

is, women's improving economic conditions (i.e., increased income, employment, and

educational attainment) would make women less dependent on males, which would threaten men

and actually cause increased violence towards females by males. For instance, Avakame (1999)

states "initial reductions in gender inequality might cause increases in violence because it would

frustrate males into intensifying their use of violence to reassert their diminishing patriarchal

power and authority" (927). Though measures of gender inequality have been identified by

feminist scholars as a cause of violence against women, research has shown conflicting findings

about the direction of the relationship. Some research has been supportive of this view (Yang and

Lester, 1988; Whaley and Messner, 2002), while others have not been (Gartner et al., 1990).

Whaley and Messner (2002) hypothesized and found that greater gender equality increased male

killings of females, results directly consistent with the backlash perspective. They concluded that

"greater gender equality is threatening to male dominance, and as such, it increases male

violence against women..." (199). In addition, other research has shown that when male

unemployment is high relative to females, wives killed by their husbands' increases (Bailey and

Peterson, 1995). However, other research shows contradictory results. Gartner et al. (1990)









Russell, Diane E. 1975. The Politics ofRape: The Victim's Perspective. New York: Stein and
Day.

Saunders, Daniel, and Angela Browne. 2000. Intimate partner homicide. In Case Studies in
Family Violence, 2nd, eds. Robert T. Ammerman and Michael Hersen. New York:
Plenum Press.

Schechter, Susan. 1982. Women andMale Violence. Boston: South End Press.

Schwartz, Martin D. 1987. Gender and injury in spousal assault. SociologicalForum 20:61-75.

Shackelford, Todd K. 2001. Partner-killing by women in cohabiting relationships and marital
relationships. Homicide Studies 5:253-266.

Shackelford, Todd K., and Jenny Mouzos. 2005. Partner killing by men in cohabiting and marital
relationships. Journal ofInterpersonal Violence 20:1310-1324.

Sherman, Lawrence, and Richard Berk. 1984. The specific deterrent effects of arrest for
domestic assault. American Sociological Review 49: 261-272.

Silverman, Robert, A., and S. K. Mukherjee. 1987. Intimate homicide: An Analysis of violence
social relationships. Behavioral Sciences and the Law 5:37-47.

Smith, M. Dwayne, and Victoria R. Brewer. 1992. A sex-specific analysis of correlates of
homicide victimization in United States cities. Violence and Victims 4:279-86.

Smith, Paige Hall, Kathryn E. Moracco, and John D. Butts. 1998. Partner homicide in context.
Homicide Studies 2:400-421.

Smith, M. Dwayne, and Margaret A. Zahn. 1998. Homicide: A Sourcebook of Social Research.
Thousand Oaks, CA: Sage.

STATA. 2003. Cross-sectional time-series reference manual. STATA Press.

Steffensmeier, Darrell, and Dana Haynie. 2000a. Gender, structural disadvantage, and urban
crime: Do macrosocial variables also explain female offending rates? Criminology
38:403-438.

Steffensmeier, Darrell, and Dana Haynie. 2000b. The structural sources of urban female violence
in the United States. Homicide Studies 4:107-134.

Steffensmeier, Darrell, and Cathy Streifel. 1992. Time-series analysis of the female percentage
of arrests for property crimes, 1960-1985: A test of alternative explanations. Justice
Quarterly 9:77-103.









Williams, Kirk R., and Robert L. Flewelling. 1987. Family, acquaintance, and stranger homicide:
Alternative procedures for rate calculations. Criminology 25:543-560.

Williams, Kirk R., and Robert L. Flewelling 1988. The Social Production of Criminal Homicide:
A Comparative Study of Disaggregated Rates in American Cities. American
Sociological Review 53:421-431.

Williams, Joyce E., and Karen A. Holmes. 1981. The SecondAssault: Rape andPublic Attitudes.
Westport, CT: Greenwood.

Wilson Martin. 1989. Spousal conflict and violence in evolutionary perspective. In Sociobiology
and the Social Sciences, eds. Robert W. Bell and Nancey J. Bell. Texas Tech
University Press.

Wilson Margo, and Martin Daly. 1992a. Till death do us part. In Femicide: The Politics of
Woman Killing, eds. Jill Radford and Diana E. H. Russell. Boston MA: Twayne.

Wilson, Margo, and Martin Daly. 1992b. Who kills whom in spouse killings? On the exceptional
sex ratio of spousal homicides in the United States. Criminology 30:189-216.

Wilson, Margo, Martin Daly, and Christine Wright. 1993. Uxoricide in Canada: Demographic
risk patterns. Canadian Journal of Criminology 35:263-291.

Wilson, Margo, Holly Johnson, and Martin Daly. 1995. Lethal and nonlethal violence against
wives. Canadian Journal of Criminology 37:331-361.

Wolfgang, Marvin E. 1958. Patterns of Criminal Homicide. Philadelphia: University of
Pennsylvania.

Yang, Bijou, and David Lester. 1988. The participation of females in the labor force and rates of
personal violence (suicide and homicide). Suicide andLife-Threatening Behavior
18:270-78.









male and female intimate partner homicide. Furthermore, results suggest that an increase in the

number of shelters per 100,000 females is related to an increase in male-perpetrated intimate

partner homicide in 1990 and 2000 and female-perpetrated intimate partner homicide in 2000. This

is a main domestic violence service measure that has historically been used to measure domestic

violence services and was expected to have an exposure reduction effect on intimate partner

homicide perpetration. Instead, it appears that this measure is associated with more males and

females reacting out in a lethal manner towards their intimate partner, whether it is for protection

or to maintain control and dominance. Dugan et al. (2003) also found a backlash effect for one of

their measures of domestic violence services. Specifically, they found that prosecutors' willingness

to prosecute violators of protection orders was related to an increase in homicides of married and

unmarried white females and unmarried African-American males. They concluded that, although

prosecutors' willingness to prosecute is advantageous, if this willingness comes without also being

able to provide adequate protection for victims, homicide may result. As stated earlier, if these

services aren't being utilized by females, they will be less likely to have an exposure reduction

effect on intimate partner homicide. Also, given the fact that the likelihood of intimate partner

homicide is extremely high when the relationship is ending, domestic violence resources need to

provide adequate protection when the relationship is ending or all the good intentions of increasing

shelter availability will not prevent intimate partner violence and homicide.

Indicators measuring the backlash perspective do not predict male-perpetrated intimate

partner homicide in 1990 or 2000 as was hypothesized. Interestingly, backlash indicators do reach

significance in the 2000 female-perpetrated intimate partner homicide model. However, the

relationship between these variables and female-perpetrated intimate partner homicide is complex.









variables on both male- and female-perpetrated intimate partner homicide over time. This is

something that has not been done in the literature to date. The time period covered includes a time

in history where dramatic changes were occurring in domesticity, the status of females, economic

deprivation and marginalization, and the availability of domestic violence resources with the

enactment of the Violence of Against Women Act in 1994.

A main finding of this research speaks to the ineffectiveness of domestic violence services.

The number of shelters per 100,000 females is significantly related to intimate partner homicide in

1990 and 2000 and over time. Despite all the efforts to increase shelter availability to females in

violent relationships, it appears that the increase in availability is actually associated with an

increase in intimate partner homicide. For instance, in 1990 and 2000 the increase in the shelter

rate was related to an increase in male-perpetrated intimate partner homicide. Furthermore, not

only was this finding seen in 2000 for female-perpetrated intimate partner homicide, this finding

was witnessed over time as well. Efforts to prevent domestic violence and homicide need to make

sure that they also provide adequate protection during times that are characterized by increased

violence.

This research adds to the knowledge of the decline in female-perpetrated intimate

homicide, but still leaves explanations behind the decline in male-perpetrated unclear. Research

needs to continue to examine gender-specific intimate partner homicide, but particularly male-

perpetrated intimate partner homicide. Surprisingly, none of the three theoretical perspectives

were associated with male-perpetrated intimate partner homicide over time, whereas some of the

predictors were associated with female-perpetrated intimate partner homicide over time. It is

extremely important to gain a better understanding of predictors of male-perpetrated intimate

partner homicide, given the fact that male-perpetrated intimate partner homicide has not decreased









The backlash hypothesis proposes that the increased status of females threatens males'

control and dominance over females and will result in more intimate partner homicide, because

males will use more violence to dominate and control their female partners. Factors that have

been thought to produce backlash effects overlap with some of the factors that are considered to

produce exposure reduction effects. Some of the factors considered include: greater economic,

employment, and educational status of females and the availability of domestic violence services

when they fail to adequately reduce contact between violent partners (Dugan et al., 1999; 2003).

Research supports these two main theoretical perspectives to explain intimate partner

homicide (Dugan et al., 1999; 2003). However, the contradictory results from these studies have

frustrated advocates. For instance, Dugan et al. (1999) reported an exposure reduction effect of

domestic violence resources on male intimate partner homicide victimization, but not on female

intimate partner homicide victimization, suggesting that the movement to prevent domestic

violence only really made men safer and not women. In addition, Dugan et al. (2003) reported

that some domestic violence services actually create a backlash effect and seem to put some

women in more danger of intimate partner homicide victimization (i.e., prosecutors' willingness

to prosecute). Both of these perspectives need to be examined more closely to determine which

factors are truly influencing gender-specific intimate partner homicide.

Due to the contradictory findings from the above two perspectives it is important to also

consider other factors. This study draws from the economic deprivation and marginalization

literature to gain a better understanding of intimate partner homicide. Economic deprivation and

marginalization have been shown to explain homicide in general and therefore seems reasonable

to include in the present research. Ideas behind economic deprivation and marginalization come

from both the strain and feminist literatures. Strain theorists relate the opportunity structure to









deprivation and marginalization, which has shown to be very important in the homicide literature

(Bailey, 1984; Blau and Blau, 1982; Land, McCall, and Cohen, 1990; Parker, McCall, and Land,

1999; Steffensmeier and Haynie, 2000a; 2000b; Whaley and Messner, 2002), in an attempt to

gain a better understanding of the differential gender-specific trends in intimate partner

homicide. Structural factors and the availability of domestic violence services are of particular

importance.

This dissertation begins with Chapter 2 detailing the three theoretical perspectives that

are of interest in this study. In Chapter 3 the data and methodology are discussed. Chapter 4

provides the results of the multivariate analysis. Chapter 5 contains conclusions and directions

for future research.









(Total number of police officers)l
1x 1,000
(Total population in 1990/2000) 1

Log Percent Hispanic. Another control variable is the percent Hispanic population. This

variable is included in the models to account for the fact that research has shown that Hispanics

are less likely to be involved in intimate partner homicide (Paulsen and Brewer, 2000). In their

analysis of spousal sex ratios of killing (i.e., the number of female intimate partner homicide

perpetrators for every 100 male intimate partner homicide perpetrators), Paulsen and Brewer

(2000) found that the largest disparities between male-perpetrated intimate partner homicide and

female-perpetrated intimate partner homicide was for Hispanics. One reason could be that

domestic violence and homicide is viewed differently in the Hispanic culture. To calculate the

percent Hispanic the total Hispanic population was divided by the total population and multiplied

by 100 to obtain a percent. The natural logarithm of the percent Hispanic was then calculated due

to the skewed nature of the independent variable.

Ln (Total Hispanic population) x100
nI (Total Population)

South. A dummy coded variable for the south is used in the present analysis to control for

any regional differences in intimate partner homicide offending. Research has shown that

homicide rates are higher in the south (Nisbett and Cohen, 1996) and that partner homicide is

higher in the south as well (Browne and Williams, 1989; Nisbett and Cohen, 1996). This

measure is based on Census Bureau definitions of southern regions (i.e., coded as 1 for cities

located in the southern region and 0 for cities not located in the southern region).

Missing Data

Missing information in the homicide data was problematic. For the most part this is

caused by the fact that participation in the Supplemental Homicide Reports program is









unemployment, and public assistance income decreased for both males and females. Moreover, the

availability of shelters and legal services for females increased, whereas the availability of male

batterers counseling programs and referral services decreased. Any of these changes may have an

impact on the gender-specific trends in intimate partner homicide that have been witnessed over

the last couple of decades.

Exposure Reduction

Some of the structural predictors are related to intimate partner homicide perpetration.

According to the exposure reduction perspective, it was hypothesized that measures that reduced

the contact between violent partners would be associated with female intimate partner homicide

perpetration in both 1990 and 2000. For 1990, cross-sectional analyses show that the exposure

reduction measure of female percent divorced is in fact associated with female-perpetrated intimate

partner homicide; however the relationship is in the opposite direction of what was predicted.

Instead of the percent divorced being associated with a decrease in female-perpetrated intimate

partner homicide, as the exposure reduction perspective would predict, the percent of females

divorced is associated with an increase in female-perpetrated intimate partner homicide. Other

research has found similar results (Dugan et al., 1999; 2003). One possible explanation may have

to do with the dynamics that come with the ending of a violent relationship. Research has

suggested that the most dangerous time in a violent relationship is when the relationship is coming

to an end (Campbell, 1992; Goetting, 1995). Some females may be reacting to male partners'

violence over the end of the relationship and may resort to lethal means to protect themselves.

None of the other exposure reduction measures were related to female-perpetrated intimate partner

homicide in 1990.

In 2000, two of the domestic violence service variables reach significance for female-

perpetrated intimate partner homicide, however in opposite directions. The number of shelters per





















9 -
0t Female-Perpetrated Intimate Partner
I Homicide
8


C 7

IX
h 6

I 1


SX I
= 4

3 3







1




1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001

Year



Figure 1-2.Gender-specific intimate partner homicide rates for large cities, 1989-2001.









Economic Deprivation and Marginalization Predictions

Although economic deprivation and marginalization is talked about differently in the

strain and feminist literatures, it is still applicable to the current study and should have an impact

on gender-specific intimate partner homicide. Economic deprivation is expected to increase both

male- and female-perpetrated intimate partner homicide, either in terms of differential

opportunity structures or marginalization.

Summary

In summary, three theoretical perspectives are examined to determine their ability in

explaining male- and female-intimate partner homicide. Each theoretical perspective is believed

to have a unique influence on male- and female-perpetrated intimate partner homicide, based on

findings from prior literature. The exposure reduction perspective and the backlash perspective

have shown to be important in understanding intimate partner homicide. Economic deprivation

and marginalization has shown to be an important explanation of homicide offending, and is

assumed to have an influence on intimate partner homicide as well. Many of the factors believed

to represent the theoretical perspectives overlap in many ways. For instance, in prior research

inequality has been used to represent economic deprivation and/or marginalization (Whaley and

Messner, 2002; Reckdenwald and Parker, 2008), as well as exposure reduction (Dugan et al.,

1999); however in this study inequality represents a measure of backlash. I have done my best to

break up the variables based on ideas from other research as well as my research hypotheses for

this study.

Hypotheses

Based on the previous literature on intimate partner homicide and homicide in general, it

is expected that indicators of exposure reduction, backlash, and economic deprivation and

marginalization will significantly impact intimate partner homicide for both males and females
















35.00%


- -


a m m


f f l rll
~IaiL


- Male
- Female


1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001


Year


Figure 1-3. Proportion of all homicides involving intimate partners, 1989-2001.


30.00%


25.00%




, 20.00%




15.00%




10.00%


5.00%




0.00%









male to female median income, the percent of males aged 16 and over who were employed

relative to the percent of females aged 16 and over who were employed, the percent of males

employed in the labor force, the percent of managers, executives, and administrators who were

male, and the ratio of males aged 25 or more with 4 or more years of college education to

females aged 25 or more with 4 or more years of college education. They measured economic

deprivation as the percentage of black males, percentage of black females, percentage of poor

males, percentage of poor females, percentage of unemployed males, percentage of unemployed

females, and the gini index of income inequality10. Results suggested that gender equality is

positively related to rates of males killing other males and males killing females in southern

cities. In addition, they found that economic deprivation had a significant positive relationship

with females killing males, females killing other females, males killing females, and males

killing other males.

Consistent with this research, other research has found that poverty and economic

inequality influence male and female homicides (Gartner, Baker, and Pampel, 1990; Smith and

Brewer, 1992). Also, research has shown that married females are at a higher risk of intimate

partner homicide victimization in cities where there are high levels of educational attainment and

employment inequalities between men and women (Bailey and Peterson, 1995). Furthermore,

research has found that women's economic power is significantly related to their involvement in

intimate partner homicide. Specifically, it has been shown that as women's economic resources

increase their involvement in intimate partner homicide decrease (Gauthier and Bankston, 1997).







10 The gini index represents an index of income inequality. The index can range from zero, representing perfect
equality, to 1, representing perfect inequality.









completely voluntary. Therefore, some law enforcement agencies fail to report their homicide

incidents to the FBI. According to the Bureau of Justice Statistics (2006) the SHRs are just over

90% complete. Though the coverage is high, there are still a number of homicides that go

unaccounted for. This underreporting by law enforcement agencies was corrected with an

adjustment factor based on the total number of homicide incidents reported to the Uniform Crime

Reporting program (see Fox, 2004 for discussion on weight calculation; Williams and

Flewelling, 1987). That is, SHR records were adjusted so that State and national total counts of

murder and nonnegligent manslaughter matched UCR estimates. It is important to note that this

weighting process assumes that the missing records are not systematically different from those

that have been reported to the FBI. In addition, there is a problem with using offender data

because there is a growing problem with unsolved murders. Ignoring homicides with missing

offender information understates homicide offending. A weighting strategy based on available

information about the victims murdered in both solved and unsolved homicides is used in the

present study to adjust for missing offender data (see Fox, 2004 for discussion). In the current

analysis all cases are weighted to reflect UCR estimated U.S. homicide counts and imputed

offender data.

Missing data were also an issue with the domestic violence service data. There were

instances where a program in a city did not report whether or not a particular service was offered.

In these cases, the mean of that particular service across cities was imputed. For the domestic

violence service variables included in the model, a dummy variable for whether the mean was

imputed for a city was included in the final models (i.e., 1 = mean imputation, 0 = no mean

imputation for a city) to determine if the missing values were missing at random. Results showed










Table 4-3. Summary of cross-sectional results, 1990 and 2000.
1990


Female Model



% divorced (+)


Male Model



% divorced (+)
# shelters (+)
# legal services (-)


Female Model



# shelters (+)
# legal services (-)


Male Model



% divorced (+)
# shelters (+)
# legal services (-)


M/F income (-)
M/F education (+)
M/F employ (-)


EcoDep Index (+)


EcoDep Index (+)


Control NONE South (+) Officer Rate (-) NONE


Exposure
Reduction


2000


Backlash


NONE


NONE


Economic
Deprivation


NONE


NONE


NONE


Control


NONE


South (+)


Officer Rate (-)


NONE









Table 3-2 reports the paired sample t-test values to show whether the variable means are

significantly different between 1990 and 2000. All the variable means are significantly different

except for the mean of the number of referral services per 100,000 females. In summary,

statistics show that both male- and female-perpetrated intimate partner homicide decreased from

1990 to 2000. In addition, for both males and females, economic deprivation and the percent

divorced decreased from 1990 to 2000, as well as the ratio of male to females' educational

attainment, the ratio of male to female employment, and the rate of male batterers counseling

programs. In contrast, the ratio of male to female median income, percent of unmarried

households, the rate of the number of shelters, the rate of the number of legal services, percent

Hispanic, residential mobility, and the officer rate increased from 1990 to 2000.

Analytical Plan

Cross-Sectional Analyses

Cross-sectional regression models are used to analyze the impact of the theoretical

concepts at two points in time (1990 and 2000). Poisson regression models will be used for the

current study, because the dependent variables are based on discrete counts of rare events and

have a skewed distribution. In addition, the dependent variables include a number of cities with

zero counts, all of which make OLS inappropriate to use (Osgood, 2000; Osgood and Chambers,

2000).

The poison regression model is the simplest regression model for count data. However, in

the poison regression model the variance is forced to equal the mean, which is often not the case

with count data. After examining the mean and variance of the current distribution it is clear that

the mean and the variance are not equal (Cameron and Trivedi, 1998). The negative binomial

regression model is more appropriate to use because it allows for overdispersion of the data (i.e.,









significantly related to both; however both are not in the predicted direction. Contrary to what

was predicted, backlash indicators influence female-perpetrated intimate partner homicide in

2000, but are not related to male-perpetrated intimate partner homicide in 2000. Also opposite of

the predictions, economic deprivation and marginalization did not influence either male- or

female-perpetrated intimate partner homicide.

Pooled Cross-Sectional Time Series Analysis: Fixed Effect Estimation

Table 4-4 and 4-5 displays the results from the pooled cross-sectional Ordinary Least

Squares fixed-effects regression models. Table 4-6 summarizes the findings. Interestingly,

results suggest that the theoretical constructs explain changes in female-perpetrated intimate

partner homicide, but not male-perpetrated intimate partner homicide. Specifically, changes in

female-perpetrated intimate partner homicide counts are associated with variables used in

exposure reduction explanations, but not with indicators used in backlash explanations. That is,

cities that are characterized by increases in the percent of females that are divorced are also

characterized by decreasing counts of female-perpetrated intimate partner homicide.

Surprisingly, the only domestic violence service variable that is significant in the female-

perpetrated intimate partner homicide model is the number of shelters per 100,000 females and

the relationship between the change in this variable and the change in female-perpetrated

intimate partner homicide is in the opposite direction of what was originally predicted. Cities that

show increases in the number of shelters per 100,000 females also show increases in female-

perpetrated intimate partner homicide counts. No significant relationship is found between

changes in the percent of unmarried partner households, the number of legal services available to

females, the number of batterers counseling programs available to males, or the number of

referral services offered to females and the female-perpetrated intimate partner homicide count.










Table 3-3. Variance inflation factors for all variables included in the models (N=178).
F-perp90 M-Perp90 F-perp00 M-perp00

Percent female divorced 1.228 1.464

Percent male divorced 1.204 1.666

Percent unmarried households 1.366 1.360 2.830 2.940

Male batterers programs per 100,000 males 1.245 1.241 1.405 1.399

Shelters per 100,000 females 2.198 2.138 2.392 2.387

Legal service programs per 100,000 females 2.101 2.129 2.874 2.861

Referral services per 100,000 females 1.179 1.163 1.300 1.323

Ratio of male to female median income 1.883 1.818 2.110 2.159

Ratio of male to female education 1.580 1.550 2.080 2.120

Ratio of male to female employment 1.443 1.468 2.446 2.514

Percent female unemployment 4.887 3.834

Percent male unemployment 6.970 4.494

Percent females in poverty 5.465 4.827

Percent males in poverty 4.697 4.040

Percent of households on public assistance 5.192 4.871 4.496 4.865

Percent hispanic 1.591 1.463 2.359 2.302

Residential mobility 1.976 2.219 1.696 1.718

Officer rate per 1,000 1.796 1.845 2.098 2.056

South 1.720 1.629 2.246 2.063









better understanding of what factors are truly influencing gender-specific intimate partner

homicide, during a time that is marked by significant transformations in domestic violence

legislation and response toward domestic violence with the enactment of the Violence Against

Women's Act of 1994. This research helps to address the contradictory findings while

controlling for a number of structural factors that have shown to be important in the homicide

literature. The current study examines both the arguments of exposure reduction and backlash, in

addition to economic deprivation and marginalization to explain the observed patterns in male-

and female-perpetrated intimate partner homicide over time, something that has not been done in

the literature to date. Measures of key concepts will be collected for 2 decennial points (1990 and

2000). Supplemental homicide files for this time period and census data will be utilized. Poisson

regression models will be used to investigate which theoretical perspective is associated with

male- and female-perpetrated intimate partner homicide in 1990 and 2000. Pooled cross-

sectional fixed effect time series regression will be used to determine whether changes from

1990 to 2000 in key structural indicators influence trends in male- and female-perpetrated

intimate partner homicide during this same time period. Overall, results suggest that key

theoretical indicators do in fact influence the trends in female-perpetrated intimate partner

homicide, but not male-perpetrated intimate partner homicide.











1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
1. Female-perp 1.00
IPH


2. % Fem poverty
3. % Fem unempl
4. % HH pub assis.
5. M/F income
6. M/F education
7. M/F % employ
8. % fem div
9. % unmarried
10. Shelter rate
11. Legal serve rate
12. Male batt. rate
13. Referral rate
14. Res. mobility
15. Hispanic(log)
16. Officer rate
17. South

1. Male-perp IPH
2. % Male poverty
3. % Male unempl
4. % HH pub assis.
5. M/F income
6. M/F education
7. M/F % employ
8. % Male div
9. % Unmarried
10. Shelter rate
11. Legal Serv rate
12. Male Batt. rate
13. Referral rate
14. Res. mobility
15. Hispanic(log)
16. Officer rate
17. South


.359 1.00
.300 .782 1.00
.244 .786 .793 1.00
-.221 -.282 -.093 -.265 1.00
-.190 -.393 -.328 -.364 .429 1.00
-.003 .181 .249 .212 .220 .208 1.00
.005 -.106 -.124 -.041 -.076 .050 -.126 1.00
.003 -.020 .053 .124 -.236 -.084 -.105 -.122 1.00
-.281 .024 -.107 -.026 .118 -.039 -.085 .036 -.129 1.00
-.223 .093 .011 .026 .078 -.017 -.068 .079 -.086 .678 1.00
-.078 .052 -.048 .016 -.077 -.178 -.071 .127 -.118 .293 .316 1.00
-.115 -.048 .084 .034 .034 -.046 -.072 .025 .017 .164 -.009 .133 1.00
-.244 -.460 -.431 -.442 .046 .111 -.226 .213 .225 .028 .016 .009 .079 1.00
-.058 -.125 .110 .072 -.159 .032 .152 .109 .198 -.126 -.105 -.045 .077 .279 1.00
.347 .467 .305 .388 -.403 -.345 -.146 -.132 -.004 -.014 -.044 .039 -.049 -.449 -.167 1.00
.183 .189 .045 -.195 .014 -.083 -.041 -.189 -.234 -.008 .010 .074 -.067 -.025 -.262 .105 1.00

1.00
.260 1.00
.199 .786 1.00
.195 .766 .863 1.00
-.250 -.322 -.193 -.265 1.00
-.177 -.405 -.335 -.364 .429 1.00
.006 .192 .218 .212 .220 .208 1.00
.050 .001 .024 .029 -.094 -.009 -.152 1.00
.004 -.021 .063 .124 -.236 -.084 -.105 -.107 1.00
-.280 .011 -.053 -.026 .118 -.039 -.085 .054 -.129 1.00
-.245 .091 .050 .026 .078 -.017 -.068 .079 -.086 .678 1.00
-.039 .062 -.018 .016 -.077 -.178 -.071 .137 -.118 .293 .316 1.00
-.121 -.033 .063 .034 .034 -.046 -.072 .005 .017 .164 -.009 .113 1.00
-.227 -.389 -.537 -.442 .046 .111 -.226 .138 .225 .028 .016 .009 .079 1.00
.018 -.041 .075 .072 -.159 .032 .152 .014 .198 -.126 -.105 -.045 .077 .279 1.00
.323 .426 .444 .388 -.403 -.345 -.146 -.048 -.004 -.014 -.044 .039 -.049 -.449 -.167 1.00
.122 .151 -.160 -.195 .014 -.083 -.041 -.188 -.234 -.008 .010 .074 -.067 -.025 -.262 .105 1.00









CHAPTER 1
INTRODUCTION

Only in the area of partner homicides do women's perpetration rates approach that of
men. (Browne, Williams, and Dutton 1999: 150)


Decline in Intimate Partner Homicide

Research on intimate partner homicide1 has increased in recent years. Increased attention

toward intimate partner homicide has resulted from the dramatic decline in perpetration that has

been witnessed in these types of homicides. This decline coincided with the remarkable decline

that was observed in total homicides rates that still puzzles researchers. Beginning in the early

1990s and extending into the 21st century, Americans witnessed an unprecedented drop in

violent crime rates, specifically a drop in homicide rates. This drop was witnessed across the

U.S., but was seen in large cities2 in particular. From 1990 to 2000, homicide rates in large cities

dropped roughly 42%3. See Figure 1-1 for an illustration of the decline in total U.S. homicide

rates in large cities4. Interestingly, the decline in adult homicide rates has been recorded for all

victim-offender relationship categories, but the sharpest decline in homicides have been observed

in the family category for both males and females (Rosenfeld, 2000), with most of those

homicides generally occurring between spouses, ex-spouses, boyfriends, and girlfriends (Bureau

of Justice Statistics, 2006; Durose et al., 2005).

Statistics reveal that U.S. intimate partner homicide counts have been declining for over

two decades (Bureau of Justice Statistics, 2006; Greenfield et al., 1998). See Figure 1-2 for


1 Intimate partners include spouses, ex-spouses, or boyfriends and girlfriends.
2 Large cities have a population of 100,000 residents or more.
3 See Figure 1-4 for summary of percent change from 1990 to 2000.
4 Homicide rates per 100,000 population were calculated by dividing the homicide counts by the population and then
multiplying that number by 100,000 to obtain a rate. Denominators were obtained from the 1990 and 2000 U.S.
Census Bureau. Annual population estimates were calculated for between census years by subtracting the 1990
population from the 2000 population and then dividing that difference over time by 10 (i.e., number of years). The
annual population estimate was then subtracted from the 1990 population to obtain a population estimate for 1989
and added to 1990 and each year after to obtain a population estimate for each of the following years.









violent partner (Allard et al., 1997), but reductions in AFDC benefits would limit opportunities

for females to leave their abusive partners and live on their own; thus resulting in more

opportunities for females to kill their abusive partners. Other variables that were shown to

produce an exposure reduction effect include: marriage rates, legal advocacy, warrentless arrest

laws, and mandatory arrest laws.

Declining domesticity.

Other common factors that have been considered to produce exposure reduction effects

are increasing divorce rates, and decreasing marriage rates. Rising divorce rates would result in

fewer married couples living together and would therefore reduce the exposure between violent

couples. The same idea is behind falling marriage rates, which would reduce the exposure of

violent couples because fewer individuals would be getting married and living together.

Rosenfeld (1997) examined intimate partner homicide trends in St. Louis and found that 30% of

the decline in African American spousal homicides was attributable to falling marriage rates and

rising divorce rates.

Dugan et al. (1999) found that domesticity does in fact influence intimate partner

homicide. They found that the decline in marriage rates is related to the decline in married male

and female intimate partner victimization. That is, they found that the higher the marriage rate in

cities, the higher the rates that wives kill their husbands and husbands kill their wives.

Conversely, a high divorce rate corresponded to fewer wives killing their husbands8.

Specifically, they found that "for every reduction of 10,000 women entering marriage, 2.3

married women's lives are saved. The same decrease in marriage among men saves the lives of

3.8 married men... and... as 10,000 additional men are granted divorce, 3.8 married male

homicides are avoided" (203). Dugan et al. (2003) found similar results. Specifically, they found

8 However, declining marriage rates was related to an increase in unmarried males killing their female partners.









LIST OF TABLES


Table page

3-1 Means, standard deviations (in parentheses) for all variables, 1990 and 2000 (N=
178). 69

3-2 T-test scores for change from 1990 to 2000 (N=178). .....................................................70

3-3 Variance inflation factors for all variables included in the models (N=178). .................71

3-4 Principal components analysis after varimax rotation 1990 (N=178). ............................72

4-1 Zero-inflated negative binomial regression equations with coefficients (and Z-
Scores) for gender-specific intimate partner homicide 1990........................ ...............84

4-2 Zero-inflated negative binomial regression equations with coefficients (and Z-
Scores) for gender-specific intimate partner homicide 2000 .........................................85

4-3 Summary of cross-sectional results, 1990 and 2000.................................. ... ................ 86

4-4 Fixed-effects regression coefficients for the relationship between changes in
predictor variables and changes in logged female intimate partner homicide counts
for 143a U S cities, 1990-2000 ........................ ........ ................................. ............... 87

4-5 Fixed-effects regression coefficients for the relationship between changes in
predictor variables and changes in logged male intimate partner homicide counts for
174a U S cities, 1990-2000 ............................................ ................. .. ...... 88

4-6 Summary of change model results, 1990 to 2000...................................... .................... 89










Table 3-5. Principal components analysis after varimax rotation 2000 (N=178).
VARIABLES Female Model Male Model

Percent below Poverty .903 .856

Percent Unemployed .903 .921

Percent of Households on Public Assistance .832 .886


Eigenvalue 2.546 2.574
% Variance Explained 85.851 85.809%

Note: only factor loadings greater than 0.5000 are reported









APPENDIX C
CORRELATION MATRIX FOR EXPLANATORY AND OUTCOME VARIABLES 2000.









householder is the person, or one of the people, in whose name the home is owned, being bought,

or rented. If there is no such person in the household, any adult household member 15 years old

and over could be designated as the householder (i.e., Person 1). Households are classified by

type according to the sex of the householder and the presence of relatives. Two types of

householders are distinguished: family householders and nonfamily householders. A family

householder is a householder living with one or more individuals related to him or her by birth,

marriage, or adoption. The householder and all of the people in the household related to him or

her are family members. A nonfamily householder is a householder living alone or with

nonrelatives only.

Unmarried partner An unmarried partner is a person who is not related to the householder,

who shares living quarters, and who has a close personal relationship with the householder.

Unmarried-Partner Household An unmarried-partner household is a household that includes

a householder and an "unmarried partner." An "unmarried partner" can be of the same or of

the opposite sex of the householder. An "unmarried partner" in an "unmarried-partner

household" is an adult who is unrelated to the householder, but shares living quarters and has a

close personal relationship with the householder. An unmarried-partner household may also be a

family household or a nonfamily household, depending on the presence or absence of another

person in the household who is related to the householder. There may be only one unmarried-

partner per household, and an unmarried partner may not be included in a married-couple

household as the householder cannot have both a spouse and an unmarried partner.

Income of individuals Income for individuals is obtained by summing the eight types of

income for each person 15 years old and over. The characteristics of individuals are based on the

time of enumeration (April 1, 2000), even though the amounts are for calendar year 1999.









Male-Perpetrated Intimate Partner Homicide 1990

After examining the male-perpetrated intimate partner homicide model for 1990 it is

clear that contrary to the first hypothesis, a number of the exposure reduction measures are

related to male-perpetrated intimate partner homicide. It was hypothesized that exposure

reduction indicators would impact female-perpetrated intimate partner homicide by reducing the

contact between violent partners. However, three exposure reduction indicators are associated

with male-perpetrated intimate partner homicide. The number of legal services per 100,000

females is negatively associated with male-perpetrated intimate partner homicide. Specifically, a

one standard deviation increase in the number of legal services per 100,000 females results in a

15% decrease in male-perpetrated intimate partner homicide counts { [exp (-.578 x .289 = .846]}.

Moreover, the percent of males divorced is positively related to male-perpetrated intimate

partner homicide. That is, a one standard deviation increase in the percent of males divorced

results in an 18% increase in male-perpetrated intimate partner homicide counts { [exp (.053 x

3.12 = 1.179]}. This contradicts the ideas behind exposure reduction. One would assume that as

the exposure between partners decreased, particularly through divorce, intimate partner homicide

would decrease as well.

Furthermore, the number of shelters per 100,000 females is positively related to male-

perpetrated intimate partner homicide. That is, a one standard deviation increase in the number of

shelters per 100,000 females results in a 26% increase in male-perpetrated intimate partner

homicide counts { [exp (.828 x .276 = 1.257]}. None of the other exposure reduction indicators

reached statistical significance.

Also surprisingly, there is a lack of support for the second hypothesis. It was

hypothesized that the backlash indicators would be significantly related to male-perpetrated

intimate partner homicide, such that an increase in females' status would be related to an









poverty threshold in either 1989 or 1999 by the total number of males for whom poverty status

was determined in 1989 or 1999 and then multiplied by 100 to obtain a percent.

(No. Females for whom poverty status was determined in 1989/1999 with incomes below poverty level)100
(Total Females for whom poverty status was deterermined in 1989/1999)

Percentage of females living in poverty. The next measure is the percent of females living

in poverty. It was calculated by dividing the number of females below the officially defined

poverty threshold in 1989 or 1999 by the total number of females for whom poverty status was

determined in 1989 or 1999 and then multiplied by 100 to obtain a percent.

(No. Males for whom poverty status was determined in 1989/1999 with incomes below poverty level)100
(Total Males for whom poverty status was deterermined in 1989/1999)

Percentage of males unemployed. Another measure representing economic deprivation

and marginalization is the percent of males unemployed. This measure was calculated by

dividing the number of males 16 yrs and older who are unemployed by the total number of males

aged 16 and older and then multiplying this by 100 to obtain a percent.

(No. Males > 16 years of age that are unemployed) x100
(No. Males > 16 years of age)

Females unemployed. Another measure representing economic deprivation and

marginalization is the percent of females unemployed. This measure was calculated by dividing

the number of females 16 years and older who are unemployed by the total number of females

aged 16 and older and then multiplying this by 100 to obtain a percent.

(No. Females > 16 years of age that are unemployed) x100
(No. Females > 16 years of age)

Percentage of households on public assistance. The final measure representing economic

deprivation and marginalization is the percent of households on public assistance income. This









Ideas behind the exposure reduction perspective stem from three main findings from previous

research.

Domestic violence research shows that intimate partner homicide is commonly the result

of prolonged violence in the relationship (Browne, 1987; Campbell, 1992; Goetting, 1995;

Smith, Moracco, and Butts, 1998; Totman, 1978) or a prior history of violence (Browne, 1987;

Chimbos, 1978; Daniel and Harris, 1982; Totman, 1978). For instance, Campbell (1992)

estimated that in Dayton, Ohio 64% of female intimate partner homicide victims were abused

physically by their offender before they were killed. Also, they estimated that 79% of male

intimate partner homicide victims had abused their offender previously during the relationship.

In addition, self-report studies have shown that abused women make many attempts to try to get

outside help before actually committing the intimate partner homicide (Sherman and Berk,

1984). For instance, in reviewing police records in Detroit and Kansas City, Sherman and Berk

(1984) discovered that in 90% of the intimate partner homicide cases police had been called at

least 1 time during the 2 years prior to the homicide. Furthermore, they found that in 54% of the

intimate partner homicide cases, police had been called 5 or more times. Other research has

shown similar results (Goetting, 1995).

Moreover, research has shown that in violent relationships women are more likely to be

seriously injured (Berk et al., 1983; Browne, 1993; Brush, 1990; Crowell and Burgess, 1996;

Fagan and Browne, 1994; Langan and Innes, 1986; Schwartz, 1987; Stark, Flitcraft, and Frazier,

1979; Stets and Straus, 1990; Stout and Brown, 1995; Straus, 1993). In addition, victim-

precipitation is more likely in female-perpetrated intimate partner homicide than male-

perpetrated intimate partner homicide (Goetting, 1995; Rosenfeld, 1997; Silverman and

Mukherjee, 1987; Wolfgang, 1958). Victim-precipitation means the victim initiated the incident









reveals that rates of female-perpetrated intimate partner homicide from 1990 to 2000 had a

greater decline than male-perpetrated intimate partner homicide. It is important to determine

what factors account for this decline as well as possible reasons why the decline was not as

drastic for male-perpetrated intimate partner homicide.

Research Questions

One of the main purposes of this study is to examine gender-specific intimate partner

homicide and determine which structural correlates contribute to the observed patterns seen in

1990 and 2000. In addition, this study will examine the changes between covariates in 1990 and

2000 to determine if any of the changes can account for the changes in male- and female-

perpetrated intimate partner homicide over time.

The following research questions will be addressed:

1. What is the impact of structural measures on male- and female-perpetrated intimate

partner homicide in 1990 and 2000?

2. Will criminological theories concerning domestic violence help explain male- and

female-perpetrated intimate partner homicide in 1990 and in 2000?

3. Does the availability of domestic violence services influence intimate partner

homicide in 1990 and 2000?

4. What is the impact of change over time (1990-2000) in key structural variables on

male- and female-perpetrated intimate partner homicide trends?

Significance of Study

This study expands on the intimate partner homicide literature in several ways.

First, this study draws from three theoretical perspectives in an effort to explain intimate partner

homicide over time. The domestic violence literature has acknowledged both the importance of

reducing the exposure between violent partners as well as backlash or retaliation by male









CHAPTER 4
RESULTS

The following chapter presents the results of the cross-sectional analyses and the pooled

cross-sectional time series analyses.

Cross-Sectional Analysis

Tables 4-1 and 4-2 present the analysis from the zero-inflated negative binomial

regression equations for male- and female-perpetrated intimate partner homicide counts in 1990

and 2000. Table 4-3 summarizes the findings. Some interesting findings emerge regarding

exposure reduction, backlash, and economic deprivation and marginalization on male- and

female-perpetrated intimate partner homicide. The following section will begin by explaining the

findings from 1990 for both the female-perpetrated intimate partner model and the male-

perpetrated intimate partner model and then will end with the cross-sectional results from the

2000 models.

Female-Perpetrated Intimate Partner Homicide 1990

After examining the female-perpetrated intimate partner homicide model for 1990 it is

clear that the first hypothesis is not supported. The only exposure reduction measure that showed

slight significance was the percent of females divorced in cities (p < .1). It was found that as the

percent of females divorced in cities increases, the counts of female-perpetrated intimate partner

homicide increase as well. Specifically, a one standard deviation increase in the percent of

females divorced results in a 13% increase in female-perpetrated intimate partner homicides

{[exp (.029 x 4.24) = 1.130]}. However, this is not in the predicted direction. According to the

exposure reduction perspective, reducing the exposure between violent couples (i.e., an increase

in female % divorced) should reduce the likelihood of female-perpetrated intimate partner

homicide. Surprisingly, none of the other exposure reduction indicators, including the measures









TABLE OF CONTENTS

page

A C K N O W L E D G M E N T S ...............................................................................................................3

LIST OF TA BLES .............. ......... ....................................................... 7

LIST O F FIG U RE S ................................................................. 8

ABSTRAC T ..........................................................................................

CHAPTER

1 IN TR O D U C T IO N ............................................................................ ............. .. 11

D decline in Intim ate Partner H om icide ........................................... ................... ...... .... 11
Importance of Examining Intimate Partner Homicide.........................................................12
Overview of Theoretical Perspectives ............................ ..... ..... ....................... ..... 14
Importance of Structural Changes and Domestic Violence Awareness .............................17
Research Questions........... .......... .......... ... .. .... ...... .... .......... 20
Significance of Study ....................................................... ............ .. ............ 20
S u m m ary ................... ...................2...................2..........

2 L ITE R A TU R E R E V IE W ........................................................................ .. .......................28

Theoretical Perspectives ..................................... .................... .......... ........ .... 30
Exposure R education H ypothesis........................................................... ............... 30
F actors that R educe E exposure .............................................................. .....................32
Exposure R education Predictions ............................................................ ............37
B acklash or R retaliation H ypothesis........................................... .......................... 38
B acklash P reductions ....................... ........................................ .. ........ ......... 4 1
Economic Deprivation and M arginalization ....................................... ............... 41
Economic Deprivation and Marginalization Predictions .........................................45
Sum m ary ........................................................................... .........................4 5
H y p o th e se s ..... ..................................................................... .......................................4 5
E exposure R edu action .......... ..................................................................... ........ .. .... 46
B ack lash .................... .......... .. ...................................4 6
Economic Deprivation and M arginalization ....................................... ............... 47
Change Models ............... ......... ...........................................47

3 D A TA A N D M ETH O D O LO G Y ........................................ ............................................49

Unit of Analysis ............ .............................. ...............................49
S o u rc e s o f D ata ................................................................................................................. 5 0
H om icide D ata ......... .. .................. ................................................ 50
D om estic V violence Service D ata ......... .... .......................... .............. ........... ...... 51
Measures ........................................ .................. 52


4









APPENDIX A
GLOSSARY

Place Places, for the reporting of decennial census data, include census designated places,

consolidated cities, and incorporated places.

South Region

S.,ntth Atlantic Division:

Delaware, Maryland, District of Columbia, Virginia, West Virginia, North Carolina, South

Carolina, Georgia, Florida

East .Sitll Central Division:

Kentucky, Tennessee, Alabama, Mississippi

West .Snlt Central Division:

Arkansas, Louisiana, Oklahoma, Texas

Employed All civilians 16 years old and over who were either (1) "at work"- those who did

any work at all during the reference week as paid employees, worked in their own business or

profession, worked on their own farm, or worked 15 hours or more as unpaid workers on a

family farm or in a family business; or (2) were "with a job but not at work"- those who did

not work during the reference week, but who had jobs or businesses from which they were

temporarily absent because of illness, bad weather, industrial dispute, vacation, or other personal

reasons. Excluded from the employed are people whose only activity consisted of work around

their own house (painting, repairing, or own home housework) or unpaid volunteer work for

religious, charitable, and similar organizations. Also excluded are all institutionalized people and

people on active duty in the United States Armed Forces.

Unemployed All civilians 16 years old and over were classified as unemployed if they were

neither "at work" nor "with a job but not at work" during the reference week, were looking for









where the variance exceeds the mean) (Gardner et al., 1995; Osgood, 2000). In the negative

binomial distribution the mean and the variance are as follows:

E(Y) = u Var(Y) = / + (,/2 /k) [1/k is the dispersion parameter]

Though negative binomial models take into account overdispersion, they usually under

predict the number of zero counts in the dependent variable. Since both male- and female-

perpetrated intimate partner homicides are rare events and therefore result in a number of zero

counts across cities, zero-inflated negative binomial regression is more appropriate and will be

utilized. Furthermore, the vuong test produced a z-value of 14.55 (p = .000) for the male-

perpetrated intimate partner homicide model and a z-value of 15.05 (p = .000) for the female-

perpetrated intimate partner homicide model. A significant value means that the zero-inflated

model is a better fit of the data. Excess zero counts may cause overdispersion and zero-inflated

models take this into consideration (Cameron and Trivedi, 1998). In both models the intimate

partner homicide count was converted into an equivalent of a rate by including the logged

gender-specific population size multiplied by a factor of 3 (i.e., to account for the 3 year pooled

dependent variable counts) as an exposure variable in the model (Agresti, 1996).

Pooled Cross-Sectional Time Series Analysis: Fixed Effect Estimation

Using STATA, version 9.1, pooled cross-sectional time series analyses with fixed effect

estimation are used to estimate the impact of key theoretical constructs on changes from 1990 to

2000 in male- and female-perpetrated intimate partner homicide. An Ordinary Least Squares

fixed effects regression15 was employed as follows



15 Negative binomial fixed effects and random effects regression models were estimated, due to the large number of

zero counts and skewed distribution for the dependent variables. However, the model for the female-perpetrated

intimate partner homicide data would not allow for the estimation of the Hausman test because the data failed to









APPENDIX B
CORRELATION MATRIX FOR EXPLANATORY AND OUTCOME VARIABLE 1990.










Table 4-4. Fixed-effects regression coefficients for the relationship between changes in predictor
variables and changes in logged female intimate partner homicide counts for 143a US cities,
1990-2000
b t

Exposure Reduction
Female percent divorced -.036+ -1.74
Percent unmarried households .065 1.26
Male batterers program rate .055 .60
Shelter rate .261* 2.15
Legal service programs rate -.035 -.40
Referral services rate -.268 -.90

Backlash
Ratio of male to female education .095 .13
Ratio of male to female income -.184 -.29
Ratio of male to female employment .299 .32

Economic Deprivation
Female economic deprivation index .035* 2.37

Controls
Percent Hispanic (log) -.260+ -1.70
Residential mobility -.023 -1.03
Officer rate -.111 -.99
1990b -.030 -.10
Female dummy log zero' -.754** -5.51
Female population 15 and over (log)d -.534** -2.81
Constant 8.90** 2.65

Within R2 .537
Hausman test .000**

Notes: (a) 70 observations (35 groups) were dropped due to all zero outcomes (i.e. zero count for intimate partner
homicide for both 1990 and 2000); (b) coefficients for period effects are relative to the reference category, 2000;
(c) represents a dummy variable for imputation of zero for log of a zero count; (d) female population is multiplied
by a factor of three to account for the dependent variable multiplied by a factor of three.
+ p < .10 p<.05 ** p< .01









for domestic violence service availability, are significantly related to female-perpetrated intimate

partner homicide.

Results are consistent with the second hypothesis. None of the backlash indicators are

significantly associated with female-perpetrated intimate partner homicide in 1990 when

controlling for a number of other structural factors. It was hypothesized that as females' status in

terms of income, employment, and educational attainment increased male-perpetrated intimate

partner homicide would increase as well. Backlash indicators were not predicted to be related to

female-perpetrated intimate partner homicide.

Consistent with the third hypothesis, economic deprivation and marginalization is

significantly related to female-perpetrated intimate partner homicide in the predicted direction.

That is, the higher the female economic deprivation index in large cities, the higher the counts of

female-perpetrated intimate partner homicide. Specifically, a one standard deviation increase in

the economic deprivation index results in a 40% increase in female-perpetrated intimate partner

homicides {[exp (.028 x 11.97 = 1.398]}. Just as predicted, it was hypothesized that economic

deprivation and marginalization would have a significant impact on female-perpetrated intimate

partner homicide, such that increases in economic deprivation and marginalization would be

related to higher female-perpetrated intimate partner homicide in cities.

In addition, despite findings from previous research none of the control variables are

significant in the model. Residential mobility, the officer rate, log of the percent Hispanic, or

whether or not the city was located in the south are not significantly related to female-perpetrated

intimate partner homicide in 1990. However, all of the coefficient signs are in the predicted

direction.










Table 4-1. Zero-inflated negative binomial regression equations with coefficients (and Z-Scores)
for gender-specific intimate partner homicide 1990.

Female Model (N=178) Male Model (N=178)


Exposure Reduction
Percent divorced (gender-specific)

Percent unmarried households

Male batterers programs rate

Shelter rate

Legal service rate

Referral services rate

Backlash
Ratio of male to female education

Ratio of male to female income

Ratio of male to female employment

Economic Deprivation
Economic deprivation index (gender-specific)

Controls
Percent Hispanic (log)

Residential mobility

Officer rate

South


Constant
Maximum likelihood R-square
Log likelihood

+ p < .10 p<.05 p


.029 (1.84)+

-.024 (-.62)

-.017 (-.27)

-.015 (-.04)

-.001 (-.00)

-.060 (-.20)


.396 (-.75)

.969 (1.39)

-.514 (-.83)


.028 (3.92)**


-.057 (-1.23)

.010 (.88)

-.069 (-.75)

.255 (2.02)


-12.44**
.831
-350.368**


.053 (3.26)**

-.024 (-.82)

.028 (.59)

.828 (2.72)**

-.578 (-1.99)*

-.202 (-.80)


-.221 (-.51)

-.041 (-.10)

-.256 (-.46)


.016 (2.80)**


-.041 (-1.06)

-.002 (-.23)

-.050 (-.68)

.332 (3.13)**


-10.359**
.821
-462.997**


< .01









None of the changes in the indicators for backlash are significantly related to the change in

female-perpetrated intimate partner homicide from 1990 to 2000.

Furthermore, changes in the female economic deprivation index are significantly

associated with changes in female-perpetrated intimate partner homicide counts. This suggests

that intra-city changes in female economic deprivation were strongly associated with within-city

increases in female-perpetrated intimate partner homicide counts in the model. Cities that are

characterized by increases in the percent of females in poverty, the percent of females that are

unemployed, and the percent of households on public assistance from 1990 to 2000 are also

characterized by increases in female-perpetrated intimate partner homicide counts.

Overall, there is mixed support for the changes in the theoretical perspectives on the

changes in female-perpetrated intimate partner homicide counts from 1990 to 2000. A change in

economic deprivation within cities is significantly associated with changes in female-perpetrated

intimate partner homicide counts. Also, changes in two indicators of exposure reduction were

related to changes in female-perpetrated intimate partner homicide counts; however one was in

the opposite direction of what was predicted.

Interestingly, support is not found for the changes in the variables identified by the

theoretical perspectives on the changes in male-perpetrated intimate partner homicide counts

from 1990 to 2000. Male- and female-perpetrated intimate partner homicides are unique events

with different motives that require a unique set of explanations. From this analysis, it appears

that changes in factors that influence female-perpetrated intimate partner homicide from 1990 to

2000 have no influence on the perpetration of male intimate partner homicide from 1990 to 2000.

Research examining these trends has acknowledged the greater support among certain factors in

explaining female-perpetrated intimate partner homicide. For instance, Dugan et al. (1999) find









purpose of capturing the three theoretical perspectives of exposure reduction, backlash, and

economic deprivation and marginalization.

Exposure reduction measures

To assess whether domesticity and domestic violence service availability had an exposure

reducing effect on intimate partner homicide, eight measures of exposure reduction are included

in this study. These measures include: percentage of divorced males, percentage of divorced

females, percentage of unmarried partner households, the number of shelters offered per 100,000

females, the number of legal services offered per 100,000 females, the number of referrals

offered per 100,000 females, and the number of male batterers counseling services offered per

100,000 males.

Percentage of divorced males. The first measure of exposure reduction is the percent of

males that are divorced. This was calculated by dividing the number of males aged 15 and older

who are divorced by the total number of males aged 15 and older. This calculation was then

multiplied by 100 to obtain a percent.

(No. Males > 15 years of age that are divorced)l 00
(No. Males > 15 years of age)

Percentage of divorced females. The percent of females that are divorced was calculated

the same as the percent of males divorced. The number of females aged 15 and older who are

divorced was divided by the total number of females aged 15 and older and then multiplied by

100 to obtain a percent.

(No. Females > 15 years of age that are divorced) x100
(No. Females > 15 years of age)

Percentage of unmarried partner households. Another measure of exposure reduction is

the percent of unmarried partner households. This measure captures cohabitation by unmarried









that indicators of exposure reduction primarily influence male intimate partner homicide

victimization.









Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy


EXAMINING CHANGES IN MALE AND FEMALE INTIMATE PARTNER HOMICIDE
OVER TIME, 1990-2000

By

Amy Reckdenwald

August 2008

Chair: Lonn Lanza-Kaduce
Cochair: Karen F. Parker
Major: Criminology, Law and Society

Research on intimate partner homicide has increased in the recent years. This is partially

due to the dramatic decline witnessed over the last couple of decades in these types of homicides

as well as the growth that has occurred in public awareness and policy responses toward

domestic violence. Recent intimate partner homicide research has predominately focused around

two perspectives to explain the relationship between intimate partner homicide and domestic

violence resources the exposure reduction hypothesis and the backlash or retaliation

hypothesis, with results that support both (Dugan, Nagin, and Rosenfeld, 1999; 2003). The

exposure reduction hypothesis proposes that domestic violence resources that reduce the

exposure or contact between intimate partners should decrease the probability of intimate partner

homicide, while the backlash hypothesis suggests the opposite. That is, domestic violence

interventions may have unintended consequences and increase the risk of intimate partner

homicide if they threaten male dominance and control over their partners. The contradictory

results from these studies have frustrated advocates and have made them question their efforts to

make females safer. Societal remedies to lower intimate partner homicide seem to be addressing

male-perpetrated intimate partner violence and homicide insufficiently. It is important to gain a









A one standard deviation increase in any of these three ratio measures results in a significant

percentage change in the female-perpetrated intimate partner homicide in 2000. An increase in the

ratio of median income and the ratio of male to female employment has a negative effect on

female-perpetrated intimate homicide (i.e., decreases), while an increase in the ratio of male to

female education has a positive effect (i.e., increases) on female-perpetrated intimate partner

homicide.

Changes Over Time

In regards to changes over time, changes in structural variables from 1990 to 2000 were

related to intimate partner homicide perpetration over time, however this is only true for female-

perpetrated intimate partner homicide. Interestingly, although many of the structural predictors

influence male-perpetrated intimate partner homicide cross-sectionally, these perspectives did not

explain male-perpetrated intimate partner homicide over time at all. Results suggest that cities that

are characterized by changes in the percent of females divorced from 1990 to 2000 are

characterized by changes in female-perpetrated intimate partner homicide counts. Also, cities

that show changes in the number of shelters per 100,000 females from 1990 to 2000 also show

changes in female-perpetrated intimate partner homicide counts. Furthermore, cities that are

characterized by changes in the percent of females in poverty, the percent of females that are

unemployed, and the percent of households on public assistance from 1990 to 2000 are also

characterized by changes in female-perpetrated intimate partner homicide counts. Lastly, cities

that are characterized by changes in the logged percent Hispanic population from 1990 to 2000

are characterized by changes in female-perpetrated intimate partner homicide counts.

Overall, there is mixed support for the changes in the theoretical perspectives on the

changes in female-perpetrated intimate partner homicide counts from 1990 to 2000 and

absolutely no support on the changes in male-perpetrated intimate partner homicide counts.












Exposure Reduction
% Divorced
% Unmarried Households
Shelter Rate
Legal Assistance Rate
Batterers Counseling Rate
Referral Rate



Economic Deprivation
% Living in Poverty
% Unemployed
% HHs on Public Assistance








1+


Female-Perpetrated
Intimate Partner Homicide





Male-Perpetrated
Intimate Partner Homicide


Figure 2-1.Detailed theoretical and conceptual model on exposure reduction, backlash, economic
deprivation and marginalization and male- and female-perpetrated intimate partner
homicide.


Backlash (higher levels of
equality)
Ratio F/M Median Income
Ratio F/M Education
Ratio F/M Employment









S(Male median income
(Female median income)

Ratio of the percentage of male employment to the percentage of female employment.

The second backlash measure is the ratio of the percentage of male employment to the

percentage of female employment. This measure was calculated by dividing the percent of males

aged 16 or older who were employed in the civilian labor force by the percent of females aged 16

or older who were employed in the civilian labor force.

(% of males aged 16 or more who are employed in the civilain labor force)
(% of females aged 16 or more who are employedin civilian labor force)

Ratio of the percentage of males with college education to the percentage of females with

college education. The final backlash measure is the ratio of the percentage of males with a

college education to the percentage of females with a college education. This measure was

calculated by dividing the percent of males aged 25 and older with 4 or more years of college

education by the percent of females aged 25 and older with 4 or more years of college education.

(% of males aged 25 and older with 4 or more years of college education)
(% of females aged 25 and older with 4 or more years of college education)

Economic deprivation and marginalization measures

Two gender-specific measures were included in the models to capture economic

deprivation and marginalization. These measures included: percentage of males living in

poverty, percentage of females living in poverty, percentage of males unemployed, and the

percentage of females unemployed. A third measure of the percentage of households on public

assistance was also included to measure economic deprivation.

Percentage of males living in poverty. The first measure is the percentage of males living

in poverty. It was calculated by dividing the number of males below the officially defined









educational attainment (i.e., gap widens) results in a 22% increase in female-perpetrated intimate

partner homicide counts {[exp (2.042 x .098 = 1.222}.

Not supporting the third hypothesis, the economic deprivation index is not significant in

the 2000 model. This is surprising considering that the economic deprivation index was

significant at the .001 level in the 1990 model (p < .001). However, examining descriptive

statistics does show a significant decrease in the female economic deprivation index from 1990

to 2000. The lack of significance in the economic deprivation index in 2000 may be due to the

differences in the U.S. economy in the 1980s and 1990s. The U.S. experienced a recession in the

late 1980s until the early 1990s. This was followed by a substantial period of economic growth

throughout the 1990s.

The only control variable that is significant in the female-perpetrated model is the officer

rate. That is, the higher the officer rate per 1,000 persons, the lower the count of female-

perpetrated intimate partner homicide. This suggests that more police officers would be able to

react quicker to domestic violence calls and therefore prevent many intimate partner homicides.

Male-Perpetrated Intimate Partner Homicide 2000

The main difference between the male-perpetrated intimate partner homicide model in

2000 compared with 1990 is that the economic deprivation index is not significant. This is

surprising, considering that economic deprivation was significant at the .001 level in 1990

(p<.001). As stated earlier, the lack of significance in the economic deprivation index in 2000

may be due to the differences in the U.S. economy in the 1980s and 1990s. However, all other

results are similar.

Of the exposure reduction indicators, the percent of males that are divorced in cities is

positively related to male-perpetrated intimate partner homicide. That is, the higher the percent

of males divorced in cities, the higher the count of male-perpetrated intimate partner homicide.


































2008 Amy Reckdenwald









Johnson, David R. 1995. Alternative methods for the quantitative analysis of panel data in family
research: Pooled time-series models. Journal ofMarriage and the Family 57:1065-
1077.

Junger-Tas, Josine and Ineke Haen Marshall. 1999. The self-report methodology in crime
research. In Crime and Justice: A Review of the Research, ed. Michael Tonry.
Chicago: University Press.

Jurik, Nancy C., and Russ Winn .1990. Gender and homicide: a comparison of men and women
who kill. Violence and Victims 5:227-242.

Kaukinen, Catherine. 2004. Status compatibility, physical violence, and emotional abuse in
intimate relationships. Journal ofMarriage and Family 66:452-471.

Land, Kenneth C., Patricia L. McCall, and Lawrence E. Cohen. 1990. Structural covariates of
homicide rates: Are there any invariances across time and social space? American
Journal of Sociology 95:922-963.

Langan, Patrick A., and Christopher A. Innes. 1986. Preventing Domestic Violence Against
Women. Washington, DC: U.S. Department of Justice, Bureau of Justice Statistics.

Maddala, G. S. 1983. Limited-dependent and Qualitative Variables in Econometrics. New York:
Cambridge University Press.

Makepeace, James M. 1989. Dating, living together, and courtship violence. In Violence in
Dating Relationships: Emerging Social Issues, eds. Maureen. A. Pirog-Good and Jan
E. Stets. New York: Praeger.

Makepeace, James M. 1997. Courtship violence as process: A developmental theory. In Violence
Between Intimate Partners: Patterns, Causes, and Effect, ed. Albert P. Cardarelli.
Boston: Allyn & Bacon.

McCall, Patricia L., and Karen F. Parker, K. F. 2005. A dynamic model of racial competition,
racial inequality, and interracial violence. Sociological Inquiry 75:273-293.

Merton, Robert K. 1968. Social Theory and Social Structure. New York: Free Press.

Messner, Scott. F., and Richard Rosenfeld. 1999. Social structure and homicide: Theory and
research. In Homicide Studies: A Sourcebook of Social Research, eds. M. Dwayne
Smith and Margaret Zahn. Thousand Oaks, CA: Sage.

Messner, Steven F., Glenn D. Deane, Luc Anselin, and Benjamin Pearson-Nelson. 2005.
Locating the vanguard in rising and falling homicide rates across U.S. cities.
Criminology 43:661-696.

National Coalition Against Domestic Violence. 1991. 1991 National Directory of Domestic
Violence Programs: A Guide to Community .\liheir, Safe Home and Service
Programs. Washington, DC: National Coalition Against Domestic Violence.









Specifically, a one standard deviation increase in the percent of males divorced results in a 12%

increase in male-perpetrated intimate partner homicide counts {[exp (.058 x 2.03 = 1.124]}. This

follows the findings from the 1990 male-perpetrated model.

Also, the number of shelters per 100,000 females is positively related to male-perpetrated

intimate partner homicide. That is, the higher the number of shelters per 100,000 females in

cities, the higher the count of male-perpetrated intimate partner homicide. Specifically, a one

standard deviation increase in the number of shelters per 100,000 females results in a 20%

increase in male-perpetrated intimate partner homicide counts { [exp (.202 x .909 = 1.202]}.

Though contradictory to the exposure reduction perspective, it is not all that surprising

considering that the same finding showed up in the 1990 male-perpetrated model.

Moreover, the number of legal services per 100,000 females is negatively related to male-

perpetrated intimate partner homicide. Specifically, the higher the number of legal services per

100,000 females in cities, the lower the count of male-perpetrated intimate partner homicide.

That is, a one standard deviation increase in the number of legal services per 100,000 females

results in a 25% decrease in male-perpetrated intimate partner homicide counts { [exp (-.244 x

1.167 = .752]}. This too, was a significant finding in 1990.

Again surprisingly, none of the backlash indicators are related to male-perpetrated

intimate partner homicide in 2000. It appears that measures of females' status relative to males'

are not predictive of male-perpetrated intimate partner homicide. Also, none of the control

indicators are significant predictors of male-perpetrated intimate partner homicide.

In summary, many of the structural indicators influence male- and female-perpetrated

intimate partner homicide. Exposure reduction measures influence both male- and female-

perpetrated intimate partner homicides, with two domestic violence service variables









subdivisions, Hawaiian home lands, Congressional districts, and ZIP Code tabulation areas. Data

are available down to the block level for many tabulations, but only to the census tract level for

others. Available on CD-ROM, DVD, and American FactFinder.

Summary File 3 This Census 2000 file presents data on population and housing long-form

subjects, such as income and education. It includes population totals for ancestry groups. It also

includes selected characteristics for a limited number of race and Hispanic or Latino categories.

The data are available for the U.S., census regions, census divisions, states and statistically

equivalent entities, counties and statistically equivalent entities, county subdivisions, places,

census tracts, block groups, metropolitan areas, urban areas, American Indian and Alaska Native

areas, tribal subdivisions, Hawaiian home lands, Congressional districts, and ZIP Code

tabulation areas. Available on CD-ROM, DVD, and American FactFinder.

UCR Uniform Crime Reports









living with anyone related by birth, marriage, or adoption, then the person's own income is

compared with his or her poverty threshold.

Individuals for whom poverty status is determined Poverty status was determined for all

people except institutionalized people, people in military group quarters, people in college

dormitories, and unrelated individuals under 15 years old. These groups also were excluded from

the numerator and denominator when calculating poverty rates. They are considered neither

"poor" nor nonpoorr." (Census Glossary)

Residence 5 years ago The data on residence 5 years earlier were derived from answers to

long-form questionnaire Item 15, which was asked of a sample of the population 5 years old and

over. This question asked for the state (or foreign country), U.S. county, city or town, and ZIP

Code of residence on April 1, 1995, for those people who reported that on that date they lived in

a different house than their current residence. Residence 5 years earlier is used in conjunction

with location of current residence to determine the extent of residential mobility of the

population and the resulting redistribution of the population across the various states,

metropolitan areas, and regions of the country.

Summary File 1 This Census 2000 file presents 100-percent population and housing data for

the total population, for 63 race categories, and for many other race and Hispanic or Latino

categories. The data include age, sex, households, household relationship, housing units, and

tenure (whether the residence is owned or rented). Also included are selected characteristics for a

limited number of race and Hispanic or Latino categories. The data are available for the U.S.,

census regions, census divisions, states and statistically equivalent entities, counties and

statistically equivalent entities, county subdivisions, places, census tracts, block groups, census

blocks, metropolitan areas, urban areas, American Indian and Alaska Native areas, tribal










APPENDIX G
CORRELATION MATRICES FOR DOMESTIC VIOLENCE SERVICE VARIABLES.

1990 1 2 3 4 5 6 7
1. Shelters 1.00
2. Hotlines .906 1.00
3. Counseling .906 .995 1.00
4. Children counseling .877 .920 .922 1.00
5. Legal services .646 .723 .736 .661 1.00
6. Batterers counseling .392 .419 .420 .422 .283 1.00
7. Number of referrals .328 .413 .418 .301 .101 -.241 1.00


2000 1 2 3 4 5 6 7
1. Shelters 1.00
2. Hotlines .904 1.00
3. Counseling .900 .981 1.00
4. Children counseling .923 .946 .956 1.00
5. Legal services .851 .921 .959 .933 1.00
6. Batterers counseling .630 .702 .734 .715 .733 1.00
7. Number of referrals -.163 -.144 -.153 -.191 -.246 -.095 1.00









husbands killing wives. In addition, Browne and Williams (1993) showed that while marital

intimate partner homicide has been decreasing over the years, male-perpetrated non-marital

intimate partner homicide has actually been increasing. Browne and Williams (1993) concluded

that the rate of unmarried females being killed by their intimate partner increased significantly

during this period, while the rate that unmarried males were killed by their intimate partner

"show no clear trend during the 1976-1987 period" (87). Similarly, in Rosenfeld's (1997)

examination of intimate partner homicide victimization in California between 1987 and 1996,

rates of marital intimate partner homicide decreased while rates of non-marital intimate partner

homicide increased. Rosenfeld (1997) attributes this marked change to an increase in the

proportion of younger people who are not married. Along these lines, Riedel and Best (1998)

also examined intimate partner homicide in California for the years 1987 to 1996 and found that

intimate partner homicide is more common in common-law relationships compared with spousal

and boyfriend/girlfriend relationships. They suggest that a reason for this may be because of the

meaning that is attributed to the relationship (Makepeace, 1997) and the lack of stability

(Makepeace, 1989).

In summary, research describing recent trends in intimate partner homicide has found

some interesting results. It appears that while total intimate partner homicide is decreasing, once

one disaggregates by gender and relationship type different trends emerge. For instance, overall

patterns show that males have experienced a greater decline in intimate partner homicide

victimization than females. In addition, married intimate partner homicide has decreased, while

non-married intimate partner homicide has increased and this increase is especially pronounced

for non-married female intimate partner victimization.









that the missing values were missing at random and therefore the mean imputations were

appropriate to use.

Multicollinearity

There are two ways to assess mulitcollinearity: (1) examining bivariate correlations

between the variables and, (2) examining variance inflation factors. The bivariate correlation

matrices for all variables in 1990 and 2000 are presented in Appendix B and Appendix C. After

examination of the correlation matrices it is evident that there is collinearity and partiality among

the regressors. While there is no clear standard, researchers generally accept a correlation greater

than .500 to mean that there may be problems with mulitcollinearity. Examination shows that

there are correlations between the independent variables that are above .500. For instance, in

1990 and 2000 the percent of female unemployment and the percent of females living in poverty

are highly correlated (r =.782 and r = .816 respectively). The percent of females living in poverty

and the percent of households on public assistance are correlated (r = .786 and r = .751). The

percent of females unemployed and the percent of households on public assistance income are

also correlated (r = .793 and r = .751). Furthermore, the percent of male unemployment and the

percent of males living in poverty are highly correlated (r=.786 and r = .788 respectively). The

percent of males living in poverty and the percent of households on public assistance are

correlated (r = .766 and r = .757). The percent of males unemployed and the percent of

households on public assistance income are also correlated (r = .863 and r = .816).

To be sure that multicollinearity is not a problem variance inflation factors were

calculated. Generally it is accepted that VIFs greater than 4 represent multicollinearity (Fisher

and Mason, 1981:109). However, some have suggested that a value of 10 or greater indicates

multicollinearity (Ott and Longnecker, 2001:652; Pindyck and Rubinfeld, 1998). Variance

inflation factors for all independent variables are presented in Table 3-3. For the female-





1. Female-perp IPH
2. % Fem poverty
3. % Fem unempl
4. % HH pub assis.
5. M/F income
6. M/F education
7. M/F % employ
8. % fem div
9. % unmarried
10. Shelter rate
11. Legal serve rate
12. Male batt. rate
13. Referral rate
14. Res. mobility
15. Hispanic(log)
16. Officer rate
17. South

1. Male-perp IPH
2. % Male poverty
3. % Male unempl
4. % HH pub assis.
5. M/F income
6. M/F education
7. M/F % employ
8. % Male div
9. % Unmarried
10. Shelter rate
11. Legal serve rate
12. Male batt. rate
13. Referral rate
14. Res. mobility
15. Hispanic(log)
16. Officer rate
17. South


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
1.00
.257 1.00
.219 .816 1.00
.147 .751 .751 1.00
-.245 -.256 -.200 -.178 1.00
-.148 -.365 -.365 -.326 .476 1.00
-.124 -.201 -.182 -.122 .396 .535 1.00
.032 -.061 -.146 -.145 -.009 -.038 -.245 1.00
.067 .339 .382 .537 -.294 -.358 -.267 .209 1.00
-.224 .089 .095 .115 -.012 -.079 -.182 .010 .150 1.00
-.197 .030 .033 .072 -.092 -.058 -.168 -.008 .124 .709 1.00
-.100 .018 -.004 -.015 -.021 -.123 -.055 -.021 -.014 .347 .431 1.00
-.137 .047 .065 -.002 .109 -.035 .065 -.034 -.060 .047 -.253 .086 1.00
-.084 -.164 -.155 -.251 -.289 .004 -.122 .162 .117 .008 .058 .019 -.013 1.00
-.058 -.061 .085 .177 -.053 .061 .477 -.345 .133 -.182 -.191 -.008 .108 .153 1.00
.311 .482 .425 .353 -.424 -.224 -.372 -.011 .152 .004 .072 .000 -.103 -.210 -.282 1.00
.096 .099 .001 -.293 -.019 -.023 -.078 .070 -.523 -.115 -.121 .112 .053 .118 -.213 .107 1.00

1.00
.210 1.00
.126 .788 1.00
.151 .757 .816 1.00
-.226 -.323 -.218 -.178 1.00
-.105 -.370 -.405 -.326 .476 1.00
-.007 -.124 -.233 -.122 .396 .535 1.00
-.013 .000 .131 -.009 -.034 -.123 -.348 1.00
.048 .371 .484 .537 -.294 -.358 -.267 .302 1.00
-.327 .054 .130 .115 -.012 -.079 -.182 .136 .150 1.00
-.241 .012 .062 .072 -.092 -.058 -.168 .085 .124 .709 1.00
-.105 .003 -.010 -.015 -.021 -.123 -.055 .022 -.014 .347 .431 1.00
-.143 .040 .066 -.002 .109 -.035 .065 .039 -.060 .047 -.253 .086 1.00
-.099 -.081 -.195 -.251 -.289 .004 -.122 .065 .117 .008 .058 .019 -.013 1.00
.071 .080 .012 .177 -.053 .061 .477 -.434 .133 -.182 -.191 -.008 .108 .153 1.00
.338 .411 .435 .353 -.424 -.224 -.372 .125 .152 .004 .072 .000 -.103 -.210 -.282 1.00
.005 .026 -.154 -.293 -.019 -.023 -.078 .040 -.523 -.115 -.121 .112 .053 .118 -.213 .107 1.00










D dependent V variable ....................................................... ..... .............. 52
Independent V ariables .................. .................................... .. ........ .... 52
Exposure reduction m easures........................................................ ............... 53
B backlash m measures ............................................................... ....55
Economic deprivation and marginalization measures................... .... ........... 56
C o n tro ls ...................................................................5 8
M issin g D ata ........................................... ............................... 5 9
M u ltic o llin e a rity ............................................................................................................... 6 1
M e th o d o lo g y ..................................................................................................................... 6 3
D e scrip tiv e S statistic s ................................................................................................. 6 3
A n a ly tic a l P la n .................................................................................................................. 6 5
C ross-Sectional A analyses ...................... .... .................... ....................65
Pooled Cross-Sectional Time Series Analysis: Fixed Effect Estimation .....................66

4 R E S U L T S ..........................................................................7 4

Cross-Sectional Analysis............................................ ......... 74
Female-Perpetrated Intimate Partner Homicide 1990 .....................................74
Male-Perpetrated Intimate Partner Homicide 1990 ............................................... 76
Female-Perpetrated Intimate Partner Homicide 2000 ..................................... 77
Male-Perpetrated Intimate Partner Homicide 2000 ..................................................... 79
Pooled Cross-Sectional Time Series Analysis: Fixed Effect Estimation .....................81

5 DISCUSSION AND CONCLUSIONS .................................................90

D iscu ssion .. . ...................... ................................................................9 0
E x p o su re R edu ctio n ................................................................................................... 9 1
B ack lash .................... ......... .. ...................................9 3
Economic Deprivation and Marginalization .................................. ...............95
Summary ............ ........................... ........................................95
Changes Over Time ............... ......... ................. 97
L im station s ............. .. ............... ................. ................................... 9 8
F utu re R research ................................................................99
C o n clu sio n ......... .... ................................................. ...........................10 0

APPENDIX

A G lo ssary ......... .... .................................................. ............................10 3

B Correlation Matrix for Explanatory and Outcome Variable 1990 .............................. 109

C Correlation Matrix for Explanatory and Outcome Variables 2000. ................................1. 11

D Simplified Correlation Matrix for Explanatory and Outcome Variables 1990. ................... 113

E Simplified Correlation Matrix for Explanatory and Outcome Variables 2000. ................ 115

F Variance Inflation Factors for Simplified Models ................................... ...................117









examined females' and males' risk of homicide victimization for the period 1950 to 1980 for 18

developed democracies (i.e., 13 western European nations, Canada, the U.S., New Zealand,

Australia, and Japan) and found a relationship between higher female education and lower

female homicide. They conclude that "the fact that the relationship between higher female

education and lower female homicide increases over time suggests that the status gained from

education acts through incremental changes in policies and norms proscribing violence against

women" (608).

Attempts have been made to test the backlash perspective on other forms of violence,

such as rape (Avakame, 1999; Baron and Straus, 1987; Ellis and Beatie, 1983), however mixed

findings have resulted. For instance, Avakame (1999) examined rape data from the National

Crime Victimization Survey (1992-1994) to determine if females' labor force participation had

an increase on females' rape victimization due to backlash, but did not find support. Instead,

Avakame (1999) found that unemployed women were at a greater likelihood of being raped.

However, Ellis and Beattie (1983) showed that gender equality (i.e., in terms of male/female

differences in earnings, educational attainment, and employment) is related to an increased

likelihood of rape. Surprisingly, Baron and Straus (1987) found the opposite results more

gender equality, lower likelihood of rape.

Individual-level research has looked at status incompatibilities between husbands and

wives and the influence on physical and emotional abuse. Kaukinen (2004) examined the effect

that men and women's relative economic contributions to the family have on husband-wife

physical and emotional abuse. Status compatibility was broken down into three statuses -









women to have jobs outside of the home and pursue higher education. Due to this transformation in

gender roles, females' economic and social status has increased. Furthermore, Felson (2006)

surveyed 10,000 men and women about their spouses' behavior and concluded that males are no

more controlling than females.

Contrary to what was expected, backlash correlates are associated with female-perpetrated

intimate partner homicide in 2000. It appears that, as the gap between male and female

employment and income widens (i.e., females' status relative to males decreases), female-

perpetrated intimate partner homicide decreases. However, as the gap between males' and females'

educational attainment widens, female-perpetrated intimate partner homicide increases. Females

may react to disparities in income, employment and educational attainment just as it is predicted

that males do, just in terms of frustration and not in terms of maintaining control and dominance

in the relationship. Along these lines, females appear to be taking their frustrations about the

widening gap in educational attainment out on their male partners, but not their frustrations about

the widening gap in income or employment. Anderson (1997) suggests that it is possible that

females with lower status relative to males are more likely to engage in violence because they are

less likely to be able to leave a violent relationship. That being said, females with a lower

educational attainment status relative to males may be less likely to be able to remove

themselves from a violent situation, possibly because education affords females opportunities to

learn about services available to reduce contact to prevent violence. Anderson (1997) found that

"women with slightly less education than their male partners have almost 3 times greater odds of

perpetrating domestic violence than women partnered with men processing the same amount of

education" (664).









Measures


Dependent Variable

The Supplemental Homicide Reports (SHR) are the data source for the two dependent

variables male- and female-perpetrated intimate partner homicide counts of arrest for the years

1990 and 2000. To determine the counts of male- and female-perpetrated intimate partner

homicide for a given city and a given year, the counts were aggregated to the city-level for each

year by the offender's gender and intimate partner relationship. Consistent with other work

(Messner et al., 2005; Villarreal, 2004) arrest counts for each city are based on three year sums

(1989, 1990, 1991; and 1999, 2000, 2001) to control for any fluctuations in reporting and relative

low frequency of offending. For the current study only murders and nonnegligent manslaughter

with single victim and single offender were included. This is consistent with other research

(Flewelling and Williams, 1999; Williams and Flewelling, 1987). Also, intimate partners include

spouses, common-law spouses, ex-spouses, and boyfriends/girlfriends. Homosexual relationships

were excluded for the purpose of this study, because the male-female dynamic is of particular

importance and because of the small number of cases (Dugan et al., 2003).

Independent Variables

U.S. Bureau of Census data for the years 1990 and 2000 were used as information for the

majority of the independent variables. Police Employee (LEOKA) Data were the source of data

for the officer rate. For simplicity, in the following section the independent variables are broken

down into 4 categories of measures: exposure reduction, backlash, economic deprivation and

marginalization, and control variables. Gender-specific indicators are utilized when appropriate.

It is acknowledged that many of these measures may overlap and could possibly fall within

another category of measures. For the purpose of this study the measures are divided with the









Gardner, William, Edward P. Mulvey, & Esther C. Shaw. 1995. Regression analyses of counts
and rates: Poisson, overdispersed poisson, and negative binomial models
Psychological Bulletin 118:392-404.

Gartner, Rosemary, Kathryn Baker, and Fred C. Pampel. 1990. Gender stratification and the
gender gap in homicide victimization. Social Problems 37:593-612.

Gauthier, DeeAnn K., and William B. Bankston. 1997. Gender equality and the sex ratio of
intimate killing. Criminology 35:577-600.

Greene, William H. 2000. Econometric Analysis. Macmillin. Goetting, Ann. 1988. Patterns of
homicide among women. Journal ofInterpersonal Violence 3:3-19.

Goetting, Ann. 1995. Homicide in Families and Other Special Populations. New York: Springer.

Greenberg, Stephanie W.,William M. Rohe, and Jay R. Williams. 1982. Safe and Secure
Neighborhoods: Physical Characteristics and Informal Territorial Control in High
andLow Crime Neighborhoods. Washington, DC: National Institute of Justice.

Greenburg, David F., and Valerie West. 2001. State prison populations and their growth, 1971-
1991. Criminology 39: 615-653.

Greenfield, Laurence A., Michael R. Rand, Diane Craven, Patsy A. Klaus, Craig A. Perkins,
Cherly Ringel, Greg Warchol, Cathy Maston, and Fox, James Allan. 1998. Violence by
Intimates: Analysis ofData on Crimes by Current or Former Spouses, Boyfriends, and
Girlfriends (NCJ-167237). Washington, DC: U.S. Department of Justice.

Harris, Anthony R., Stephen H. Thomas, Gene A. Fisher, and David J. Hirsch. 2002. Murder
and medicine: The lethality of criminal assault 1960-1999. Homicide Studies 6:128-
166.

Hausman, Jerry, Bronwyn H. Hall, and Zvi Griliches. 1984. Econometric models for count data
with an application to the patents-R & D relationship. Econometrica 52:909-938.

Heimer, Karen. 2000. The nature of crime: Continuity and change. Changes in the gender gap in
crime and women's economic marginalization. Criminal Justice 1:427-483.

Heimer, Karen, Stacy M. Wittrock, and Halime Unal. 2005. Gender, crime and the economic
marginalization of women. In Gender and Crime: Patterns of Victimization and
Offending, eds. Karen Heimer and Candace Kruttschitt. New York: New York
University Press.

Hunnicutt, Gwen, and Lisa M. Broidy. 2004. Liberation and economic marginalization: A
reformulation and test of (formerly?) competing models. Journal of Research in Crime
and Delinquency 4:130-155.









intimate partner homicides decreased from 1990 to 2000. The mean of female-perpetrated

intimate partner homicides (i.e., 3 year sum of counts) in 1990 is 6.553 (C = 10.864). In

comparison, the mean of female-perpetrated intimate partner homicides in 2000 is 2.970 (a =

4.779). The mean of male-perpetrated intimate partner homicides in 1990 is 11.569 (a = 18.378)

and the mean in 2000 is 8.770 (a = 13.929).

For the indicators of exposure reduction, both the percent of males divorced and females

divorced decreased from 1990 to 2000, whereas the percent of unmarried partner households

increased. In addition, the rate of the number of shelters available to females and the rate of the

number of legal services available to females increased from 1990 to 2000 (.354 to 1.49 and .313

to 1.42, respectively). In contrast, the rate of the number of male batterer's counseling programs

and the rate of the number of referral services offered decreased from 1990 to 2000 (.746 to .552

and .089 to .075, respectively).

Examining the indicators of backlash shows that the disparities between male and female

educational attainment and employment decreased (1.33 to 1.14 and 1.24 to 1.19, respectively),

whereas the disparity between male and female income increased from 1990 to 2000 (1.42 to

1.55). This suggests that over time, the gap between males' and females' educational attainment

and employment are becoming more narrower, while at the same time the gap between males'

and females' median income is widening.

For the indicators of economic deprivation and marginalization, both the male and female

economic deprivation indices decreased from 1990 to 2000. In 1990 the mean of the female

economic deprivation index was 29.38 (a = 11.97) and in 2000 it was 15.88 (o= 6.22). In 1990

the mean of the male economic deprivation index was 26.97 (o= 11.28) and in 2000 was 14.88

(a= 5.88).









629,308), ranging from 100,217 to 7,322,564. The average city population in 2000 is 355,578 (a

= 686,124), ranging from 95,658 to 8,008,278.

Sources of Data

Multiple data sources were used for the current study. The dependent variables are based

on data from the Supplemental Homicide Files for the years 1989-1991 and 1999-2001 (Fox,

2005). U.S. Bureau of Census summary files 111 and 312 for 1990 and 2000 are the sources of

data for the independent variables. Uniform Crime Reports: Police Employee (LEOKA) Data are

the source of data for police force size (consistent with work by Steffensmeier and Haynie

2000a; 2000b). Also, the 1991 and 1999-2000 Domestic Violence Service Directory, collected

by the National Coalition Against Domestic Violence, is the source for the data on the

availability of domestic violence services. These years were chosen because they were the years

published by the National Coalition Against Domestic Violence and represented 1990 and 2000

closest.

Homicide Data

The homicide data used in this study were obtained from the Supplemental Homicide

Reports (SHR). The SHRs are part of the Uniform Crime Reporting Program (UCR). In addition

to monthly criminal offense information compiled for UCR purposes, law enforcement agencies

submit supplemental data to the Federal Bureau of Investigation on homicide incidents. The

SHRs contain detailed, incident-level data on nearly all murders and nonnegligent manslaughter

that have occurred in the United States for a given year. SHRs contain information for each





1 Summary File 1 presents 100-percent population and housing data for the total population, for 63 race categories,
and for many other race and Hispanic or Latino categories. See Appendix A for a more detailed definition.
12 Summary File 3 presents data on population and housing long-form subjects, such as income and education. See
Appendix A for a more detailed definition.









LIST OF FIGURES


Figure page

1-1 Total homicide rate per 100,000 residents for large cities, 1989-2001 ............................24

1-2 Gender-specific intimate partner homicide rates for large cities, 1989-2001 ..................25

1-3 Proportion of all homicides involving intimate partners, 1989-2001..............................26

1-4 Summary of percent change from 1990 to 2000 in total homicide rate, female-
perpetrated intimate partner homicide rate, and male-perpetrated intimate partner
h o m ic id e rate ...................................... ................................... ................ 2 7

2-1 Detailed theoretical and conceptual model on exposure reduction, backlash,
economic deprivation and marginalization and male- and female-perpetrated intimate
partner homicide. ....................... .......... ... .. .. ...... ........... 48









CHAPTER 3
DATA AND METHODOLOGY

This section details the methodological strategy used in this study. To begin, the data

sources and the unit of analysis are discussed. Next, all measures are operationalized. After

examining the data three methodological issues are apparent. First, there is a number of missing

cases in the dataset in regards to offender information and domestic violence resource information.

Techniques for dealing with this will be presented. Second, there is evidence of collinearity

between predictors included in this study. To address this concern, principal components

(specifically principal axis factoring) analysis is employed. Third, Ordinary Least Squares

regression estimations are inappropriate due to the skewed nature of the dependent variables.

Poisson-based regression techniques will be utilized for the cross-sectional models, because the

dependent variables are based on discrete counts of rare events and have a skewed distribution.

These issues will be outlined in the following pages.

Unit of Analysis

For this study, the unit of analysis is cities with a population of 100,000 or more in 1990.

This definition of urban cities is consistent with the Bureau of Justice Statistics definition of

large cities (Bureau of Justice Statistics 2006). The resulting urban sample contains 200 cities

that meet that criterion. Of those cities 178 cities were used in the analysis. Chattanooga, TN,

Citrus Heights, CA, East Los Angeles, CA, Kansas City, KS, Lowell, MA, Metairie, LA,

Omaha, NB, Overland Park, KS, Paradise, NV, Springfield, IL, Sterling Heights, MI,

Worchester, NY, and all cities in the state of Florida were dropped from the analysis due to not

reporting at all during either the period 1989-1991 or 1999-2001. Chicago, IL was the final city

dropped due to being an extreme outlier. The average city population in 1990 is 323,445 (a =









Stets, Jan E., and Murray A. Straus .1990. Gender differences in reporting marital violence and
its medical and psychological consequences. In Physical violence in American
families: Risk factors and adaptations to violence in 8,145families. Murray A. Straus
and Richard J. Gelles. New Brunswick, N.J.: Transaction.

Stout, Karen D., and Patricia Browne. 1995. Legal and social differences between men and
women who kill intimate partners. Affilia 10:194-205.

Stark, Evan, Ann Flitcraft, and William Frazier. 1979. Medicine and patriarchal violence: the
social construction of a "private" event. International Journal of Health Services
9:461-493.

Straus, Murray A. 1993. Physical assaults by wives: A major social problem. In Current
Controversies on Family Violence, eds. Richard J. Gelles and Donileen R. Loseke.
Thousand Oaks, CA: Sage.

Straus, Murray A., Richard J. Gelles, and Suzanne K. Steinmetz. 1980. Behind closed doors:
Violence in the American family. Doubleday, Garden City, NY.

Totman, Jane. 1978. The Murderess: A Psychological Study of Criminal Homicide. San
Francisco, CA: R & E Research Associates.

U.S. Bureau of the Census. 1994. Census of Population and Housing, 1990. Summary Tape File
la and 3a. Washington D.C.: U.S. Department of Commerce, Bureau of the Census,
1991.

U.S. Bureau of the Census. 2003. Census of Population and Housing, 2000. Summary Tape File
1 and 3. Washington, D.C.: U.S. Department of Commerce, Bureau of the Census,
2002.

U.S. Department of Justice, Federal Bureau of Investigation. Uniform Crime Reporting Program
Data [United States]: Police Employee (LEOKA) Data, 2000 [Computer file].
Compiled by the U.S. Dept. of Justice, Federal Bureau of Investigation. ICPSR ed.
Ann Arbor, MI: Inter-university Consortium for Political and Social Research
[producer and distributor], 2002.

Vieraitis, Lynne M., Marian R. Williams. 2002. Assessing the impact of gender inequality on
female homicide victimization across U.S. cities: A racially disaggregated analysis.
Violence Against Women 8:35-63.

Villarreal, Andreas. 2004. The Social Ecology of Rural Violence: Land Scarcity,
theOrganization of Agricultural Production, and the Presence of the State. American
Journal of Sociology 110:313-348.

Walby, Sylvia. 1986. Patriarchy at Work. Minneapolis, MN: University of Minnesota Press.

Whaley, Rachel Bridges, and Steven F. Messner. 2002. Gender equality and gendered
Homicides. Homicide Studies 6:188-210.









partners as a response or consequence to interventions designed to reduce violence between

couples. However, this research has found results that support both perspectives. This study

draws from two domestic violence perspectives while also controlling for a third theory of

economic deprivation and marginalization in attempt to more clearly understand trends in

intimate partner homicide.

This study examines both male- and female- perpetrated intimate partner homicide and

compares their trends over time. This study not only documents change but also explains

changes in intimate partner homicide trends over time and gives a better understanding of what is

correlated with this observed change. Examining changes in key indicators on intimate partner

homicide over time has not been done in the literature. The dynamic nature of key indicators will

be examined to determine whether or not the changes from 1990 to 2000 in key indicators

correlates with the change in male- and female-perpetrated intimate partner homicide over time.

For example, it is of interest to determine if the change in the percent divorced in cities from

1990 to 2000 is related to the changes in intimate partner homicide during this same time period.

Focusing on changes over time will give a better understanding of the differential trends in male-

and female-perpetrated intimate partner homicide.

The time period being examined in this study depicts a time where there has been

significant changes in homicide itself as well as domesticity (i.e., living arrangements), the status

of females, and economic deprivation and marginalization. Moreover, during this time, domestic

violence was being seen as a national social problem and funding for domestic violence services

were increasing significantly with the enactment of the 1994 Violence Against Women's Act

(VAWA). This study controls for the influence of various domestic violence services while














1. Female-perp IPH
2. Econ. dep. index
3. M/F income
4. M/F education
5. M/F % employment
6. % Fem divorced
7. % Unmarried HH
8. Shelter rate
9. Legal service rate
10. Male batt. rate
11. Referral rate
12. Res. mobility
13. % Hispanic (log)
14. Office rate
15. South

1. Male-perp IPH
2. Econ. dep. index
3. M/F income
4. M/F education
5. M/F % employment
6. % Male divorced
7. % Unmarried HH
8. Shelter rate
9. Legal Service rate
10. Male batt. rate
11. Referral rate
12. Res. mobility
13. % Hispanic (log)
14. Office rate
15. South


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
1.00
.359 1.00
-.221 -.241 1.00
-.190 -.393 .429 1.00
-.003 .209 .220 .208 1.00
.005 -.116 -.076 .050 -.126 1.00
.003 .000 -.236 -.084 -.105 -.122 1.00
-.281 -.012 .118 -.039 -.085 .036 -.129 1.00
-.223 .074 .078 -.017 -.068 .079 -.086 .678 1.00
-.078 .026 -.077 -.178 -.071 .127 -.118 .293 .316 1.00
-.115 -.012 .034 -.046 -.072 .025 .017 .164 -.009 .133 1.00
-.244 -.473 .046 .111 -.226 .213 .225 .028 .016 .009 .079 1.00
-.058 -.063 -.159 .032 .152 .109 .198 -.126 -.105 -.045 .077 .279 1.00
.347 .442 -.403 -.345 -.146 -.132 -.004 -.014 -.044 .039 -.049 -.449 -.167 1.00
.183 .156 .014 -.083 -.041 -.189 -.234 -.008 .010 .074 -.067 -.025 -.262 .105 1.00

1.00
.254 1.00
-.250 -.298 1.00
-.177 -.403 .429 1.00
.006 .210 .220 .208 1.00
.050 .016 -.094 -.009 -.152 1.00
.004 .004 -.236 -.084 -.105 -.107 1.00
-.280 -.008 .118 -.039 -.085 .054 -.129 1.00
-.245 .083 .078 -.017 -.068 .079 -.086 .678 1.00
-.039 .040 -.077 -.178 -.071 .137 -.118 .293 .316 1.00
-.121 -.005 .034 -.046 -.072 .005 .017 .164 -.009 .133 1.00
-.227 -.454 .046 .111 -.226 .138 .225 .028 .016 .009 .079 1.00
.018 -.007 -.159 .032 .152 .014 .198 .198 -.105 -.045 .077 .279 1.00
.323 .452 -.403 -.345 -.146 -.048 -.004 -.004 -.044 .039 -.049 -.449 -.167 1.00
.122 .061 .014 -.083 -.041 -.188 -.234 -.34 .010 .074 -.067 -.025 -.262 .105 1.00









Economic Deprivation and Marginalization

According to the ideas behind economic deprivation and marginalization, it was

hypothesized that measures of poverty, unemployment and the percent of the population on public

assistance would be associated with both male- and female-perpetrated intimate partner homicides.

Support for this hypothesis was found in 1990. Poverty, unemployment, and having to rely on

public assistance income may significantly impact the dynamics within the home, either in terms

of frustration or strain. Males and females may react to this strain by lashing out in a lethal

manner at their intimate partners.

Summary

The results of this analysis offer support for economic deprivation and marginalization in

explaining both male- and female-perpetrated intimate partner homicide in 1990. Results also

suggest an association between some of the variables representing the exposure reduction

perspective and male- and female-perpetrated intimate partner homicide in both 1990 and 2000. In

general, the three exposure reduction indicators that are consistently significant in 1990 and 2000

are gender-specific percent divorced and two indicators measuring the availability of domestic

violence services the shelter rate and the legal service rate. Importantly, these measures are

related to both male- and female-perpetrated intimate partner homicides in a similar fashion. A one

standard deviation increase in any of these measures results in a significant percentage change in

intimate partner homicide counts. The availability of legal services in cities supports the exposure

reduction hypotheses. That is, an increase in the number of legal services per 100,000 females is

related to an increase in male-perpetrated intimate partner homicide in 1990 and 2000 and female-

perpetrated intimate partner homicide in 2000.

However, not all of the significant exposure reduction measures influence intimate partner

homicide in the predicted direction. For instance, an increase in the percent divorced increases both











Table 3-2. T-test scores for change from 1990 to 2000 (N=178).
Mean 1990 Mean 2000 Mean Diff


Female-perpetrated intimate partner homicide

Male-perpetrated intimate partner homicide

Percent female divorced

Percent male divorced

Percent unmarried households

Male batterers counseling per 100,000 males

Shelters per 100,000 females

Legal service programs per 100,000 females

Referral services per 100,000 females

Ratio of male to female education

Ratio of male to female income

Ratio of male to female employed

Female economic deprivation Index

Male economic deprivation index

Percent hispanic

Residential mobility

Officer rate

+ p < .10 p<.05 p< .01


6.55(10.86)

11.57(18.38)

13.31(4.24)

9.92(3.12)

3.91(1.54)

.75(1.09)

.35(.28)

.31(.29)

.09(.24)

1.33(.14)

1.42(.17)

1.24(.11)

29.38(11.97)

26.97(11.28)

13.40(15.67)

50.22(7.88)

1.98(.910)


2.97(4.78)

8.77(13.93)

12.25(2.08)

9.31(2.03)

5.25(1.07)

.55(.74)

1.49(.91)

1.42(1.17)

.08 (.19)

1.14(.10)

1.55(.18)

1.19(.09)

15.88(6.22)

14.88(5.88)

18.61(18.17)

51.40(6.71)

2.22(1.04)


-3.58

-2.80

-1.06

-.61

1.34

-.20

1.14

1.11

-.01

-.19

.13

-.05

-13.5

-12.09

5.21

1.18

24


-5.83**

-4.32**

-3.45**

-2.76**

12.17**

-2.99**

18.11**

13.29**

-.695

-28.94**

12.09**

-5.97**

-13.64**

-24.85**

14.03**

2.76**

4 74**









traditional status, status parity, and status reversal.9 Kaukinen (2004) found that there was no

effect of status compatibility on physical violence, but there was an effect for emotional

violence. Specifically, women in status reversal relationships are at risk for emotional abuse by

their partner. It seems that men in relationships where their female partners had a higher status

exerted their control by emotionally abusing their partners.

Backlash Predictions

Despite the mixed results for the backlash perspective, research has suggested that power

differences between partners increase the risk for abuse among women (Anderson, 1997). In

addition, men who kill their intimate partner do so to dominate and control their female partner

(Wilson, 1989; Wilson and Daly, 1992a, 1992b). It is therefore important to include measures of

backlash in the study of male- and female-perpetrated intimate partner homicide to determine if

these factors influence intimate partner homicide in 1990 and 2000 and whether the changes in

these factors contribute to the changes in intimate partner homicide over time. It is expected that

measures of backlash will have a significant impact on male-perpetrated intimate partner

homicide, because males will use violence to exert control over their female partners in

conditions where female status is high or in conditions where they feel a perceived lack of

control over their female partners.

Economic Deprivation and Marginalization

The importance of economic deprivation as a factor influencing crime has been

documented in both the strain and feminist literatures and it is consistent with the other two

perspectives of exposure reduction and backlash. The strain literature posits opportunity

structures to be related to crime rates (Merton [1949], 1968). Merton ([1949], 1968: 185-248)


9 In traditional status the economic status between the woman and the partner favored the partner. In status parity the
economic status between the woman and the partner was equal. In status reversal the economic status between the
woman and the partner favored the woman.









gender-specific total intimate partner homicide rates for large cities, 1989-20015. Surprisingly,

male- and female-perpetrated intimate partner homicide rates appear to follow a similar pattern,

though rates of male-perpetrated intimate partner homicide are higher than female-perpetrated

rates. That is, both male and female intimate partner homicide rates fluctuate but are fairly stable

from 1989 until 1992. From 1992 to 1994 a dramatic spike in the rates of intimate partner

homicide is apparent. After 1994, intimate partner homicide rates level off. Female-perpetrated

intimate partner homicide rates remain fairly stable from 1995 to 2001, while male-perpetrated

intimate partner homicide rates decline slightly. Though, the rates of male and female intimate

partner homicide show a similar pattern over time, a greater decline is observed in the rate of

female-perpetrated intimate partner homicide. Specifically, the rate of male-perpetrated intimate

partner homicide decreased 33% from 1989 to 2001, whereas the rate of female-perpetrated

intimate partner homicide decreased 45% during this same time. The reason for the more

substantial decrease in the rate of female-perpetrated intimate partner homicide compared with

male-perpetrated intimate partner homicide is unknown and remains an interest to researchers.

Importance of Examining Intimate Partner Homicide

There are many reasons that make intimate partner homicide an important topic for

research. First and foremost, as stated earlier, intimate partner homicide has decreased in recent

years with no concrete explanation. Furthermore, females are more likely to be killed by an

intimate partner than any other offender. Among female homicide victims, roughly 1 in 3 is

killed by an intimate partner, whereas among male homicide victims, only 5% are killed by an

intimate partner. Moreover, although intimate partner homicide rates have been decreasing, the

proportion of female homicide victims killed by an intimate partner has gradually increased from

1989 to 2001 (27.7% and 31.4%, respectively), whereas the proportion of male homicide victims

5 See footnote 4 for the rate calculation.









that later resulted in their own death by their intimate partner. Rosenfeld's (1997) results support

this idea as do a number of other research studies (Campbell, 1992; Cazenave and Zahn, 1992;

Goetting, 1995; Jurik and Winn, 1990; Silverman and Mukherjee, 1987; Wilson and Daly,

1992a).

To summarize, intimate partner homicide is more likely to occur in a relationship where

there is a history of violence. In these violent relationships, women are more likely to be

seriously injured than men. Furthermore, victim-precipitation is more likely in female-

perpetrated intimate partner homicide. These findings suggest that women have very different

motives for killing their intimate partners and women are more likely to kill intimate partners out

of protection or self-defense (Bernard et al., 1982; Block and Christakos, 1995; Browne, 1985;

1987; Campbell, 1992; Daly and Wilson, 1988; Dobash et al., 1992; Dugan et al., 1999;

Goetting, 1988; Peterson, 1999; Stout, 1991; Stout and Brown, 1995; see Archer, 2000 and

Saunders and Browne, 2000 for a review). That is, when women kill intimate partners it is likely

a result of attempting to prevent their partners from hurting their children or themselves, during a

violent attack by their partner, or in an attempt to prevent a future attack (Browne, 1986; 1987;

Dugan et al., 1999). According to Wilson and Daly (1992b) "Unlike men, women kill male

partners after years of suffering physical violence, after they have exhausted all available sources

of assistance, when they feel trapped, and because they fear for their own lives" (206).

Factors that Reduce Exposure

Availability of domestic violence resources.

A main factor that has been shown to influence female-intimate partner homicide by

reducing the exposure of violent partners is the availability of domestic violence resources.

Browne and Williams (1989) was the first study to examine the influence of the availability of

domestic violence resources (i.e., number of shelters; wife abuse programs other than shelters)









and domestic violence legislation (i.e., statutes providing civil injunction relief for victims of

abuse; statutes providing temporary injunction relief during divorce, separation, or custody

proceeding; statutes defining the physical abuse of a family or household member as a criminal

offense; statutes permitting warrantless arrest based on probable cause in domestic violence

cases; statutes requiring data collection and the reporting of family violence by agencies that

serve these families; and statutes providing funds for family violence shelters) on female-

perpetrated intimate partner homicide. Following prior work on homicide and intimate partner

homicide, specifically, (Browne and Flewelling, 1986) Browne and Williams (1989) examined

the rate of intimate partner homicide in U.S. states for the years 1976-1979 and 1980-1984 and

found that indeed, domestic violence resources have the ability to provide abused females with

alternatives besides killing their male partners. Specifically, they found a negative relationship

between both domestic violence resources and domestic violence legislation and intimate partner

homicide, however once controls were included in the analysis only the domestic violence

resources remained significant. Other research has also supported this hypothesis and shown that

social response to domestic violence may be important in explaining intimate partner homicide

(Browne and Williams, 1989; Dugan et al., 1999; Dugan et al., 2003).

According to Browne and Williams (1989) domestic violence resources provide the

means for women "(1) to seek protection for themselves and their children, such as emergency

restraining orders against abuse and harassment or police removal of the abusive mate; (2) to

employ more direct avenues of escape, such as shelters where they can hide from the abuser for

short periods; (3) to utilize third-party interventions, such as support groups, crisis counseling,

and legal aid, which give advise and encouragement in identifying effective nonviolent means of

responding to threat or danger; and, in some areas, (4) to receive the benefits from court-









Changes in economic deprivation within cities are significantly associated with changes in

female-perpetrated intimate partner homicide counts. Also, changes in two indicators of

exposure reduction were related to changes in female-perpetrated intimate partner homicide

counts; however the shelter rate was in the opposite direction of what was predicted and

increased shelter availability appears to be associated with more harm than good if shelters are

not being taken advantage of or if shelters are not providing adequate protection.

Limitations

There are some limitations to this research. First and foremost, the data utilized for the

dependent variables was based on Supplemental Homicide Files. The deficiencies of the database

have been well-documented (Junger-Tas and Marshall, 1999). These include underreporting,

errors in the assignment of relationships, failure to identify relationship in a substantial portion of

cases, and other missing information SHFs include information on homicides that were submitted

by law enforcement agencies to the FBI. Submission of homicide information is completely

voluntary and therefore not 100% accurate. In addition, many homicides are coded as having an

unknown victim-offender relationship. A weighting procedure was utilized to account for this lack

of reporting and lack of relationship information.

Moreover, it is acknowledged that actual usage of the domestic violence services was

unable to be captured in cities. Just because shelters, legal services, batterers counseling programs,

and referrals are offered does not mean they are being utilized to the fullest by individuals in

violent relationships. Also, the size of the domestic violence service programs (i.e., service

capacity) was unable to be captured. If domestic violence services are not being utilized or if

programs have limited capacity, domestic violence services would have a limited effect on intimate

partner homicide.










Table 4-6. Summary of change model results, 1990 to 2000.
Female Model

Exposure Reduction % divorced (-)
# shelters (+)


Backlash


NONE


Economic Deprivation


Control


EcoDep Index (+)


% Hispanic (-)


Male Model

NONE


NONE


NONE


NONE









Though research has not come to a definitive explanation for these trends, research that

has controlled for the impact of key factors has tended to attribute the recent patterns to the

social responses toward domestic violence; specifically, the availability of domestic violence

resources (Browne and Williams, 1989; Dugan et al., 1999; Dugan et al., 2003) and have found

that the availability of legal (i.e., presence of statutes pertaining to domestic violence) and extra-

legal services (i.e., number of shelters and other programs) to be related to the decline in the rates

of female-perpetrated intimate partner homicide, but not associated with the rates of male-

perpetrated intimate partner homicide (Browne and Williams, 1989). However, research has also

shown that some domestic violence resources (i.e., prosecutors' wiliness to prosecute) that were

intended to help women have the unintended consequence of making them more at risk for

intimate partner homicide victimization (Dugan et al., 2003). Increasing divorce rates, declining

marriage rates, and an improved economic status of women have also shown to be important in

explaining intimate partner homicide (Dugan et al., 1999; 2003; Rosenfeld, 1997). These factors

will be examined more closely as I describe empirical findings for the theoretical perspectives of

interest in this study.

Theoretical Perspectives

Three theoretical perspectives are of interest to this study. The following pages will

explain each perspective the exposure reduction hypothesis, the backlash or retaliation

hypothesis, and economic deprivation and marginalization. The section will end with detailed

hypotheses for the current study.

Exposure Reduction Hypothesis

According to the exposure reduction hypothesis, limiting the exposure or contact of

intimate partners to one another should decrease the probability of intimate partner homicide,

because there is less exposure to a violent partner and more opportunities to exit the relationship.














1990 2000 % Change

Total Homicide Rate 20.9827 12.1792 41.9

Female-Perpetrated IPH Rate 1.520151 0.730003 51.9

Male-Perpetrated IPH Rate 5.066648 3.325714 34.36

Figure 1-4. Summary of percent change from 1990 to 2000 in total homicide rate, female-
perpetrated intimate partner homicide rate, and male-perpetrated intimate partner
homicide rate.









violence. Merton, ([1949] 1968: 185-248) argues that deviance is a result of an unequal

distribution of opportunities among classes in society. According to Merton, the distribution of

opportunities in the social structure determines which classes are most likely to be involved in

crime and deviance. Males or females that lack certain opportunities because of deprivation (i.e.,

poverty and unemployment) may be frustrated or strained and this strain may influence intimate

partner homicide perpetration. On the other hand, feminists scholars tend to focus on inequalities

between men and women. Feminist literature examining inequalities tend to concentrate on the

economic marginalization hypothesis, which focuses on the economic disadvantage of women

relative to men (Heimer, 2000). The economic marginalization hypothesis argues that the

economic disadvantage of women (i.e., female unemployment and poverty) is a significant

predictor of female crime (Box and Hale, 1983, 1984; Heimer, 2000; Heimer, Wittrock, and

Unal, 2005; Hunnicutt and Broidy, 2004; Steffensmeier and Streifel, 1992). Extant homicide

research has incorporated the ideas behind economic deprivation and marginalization (Bailey,

1984; Blau and Blau, 1982; Land, McCall, and Cohen, 1990; Parker, McCall, and Land, 1999;

Steffensmeier and Haynie, 2000a; 2000b; Whaley and Messner, 2002); however limited research

has included indicators of economic deprivation and marginalization to explain intimate partner

homicide (for an exception see Reckdenwald and Parker, 2008). Economic deprivation and

marginalization in terms of poverty or unemployment may influence male and female intimate

partner homicide perpetration.

All three theoretical perspectives have received support in the literature; though at times

some of the findings are mixed and even contradictory. This study hopes to determine which

theoretical factors significantly contribute to male- and female-perpetrated intimate partner









males safer instead. Divorce rates have also proved to be a significant factor influencing intimate

partner homicide; as well as cohabitation of intimate partners, and improved status of females. It

is believed that measures of exposure reduction will have a significant effect on female-

perpetrated intimate partner homicide by creating other options for females besides killing their

male partner.

Backlash or Retaliation Hypothesis

Although research has shown the importance of reducing the exposure between intimate

partners in violent relationships, it is well known that the highest risk for homicide is when the

victim leaves the relationship and this is especially true for females being killed by their male

partners (Block and Christakos, 1995; Block, 2000). Research has suggested the possibility of

retaliation by the abusive partner from domestic violence interventions (Campbell, 1992;

Goetting, 1995). Dugan et al. (2003)'s findings are supportive of this statement. Dugan et al.

(2003) found a retaliation effect where domestic violence resources actually increased homicide

between intimate partners because they failed to effectively reduce exposure between intimate

partners. In fact, the prosecutor's willingness to prosecute violators of protection orders, though

intended to reduce exposure between violent intimate partners, actually caused a retaliation

effect where homicide increased for married and unmarried white females and African-American

unmarried males. They concluded that "being willing to prosecute without providing adequate

protection may be harmful" (192).

Contrary to exposure reduction predictions, research has hypothesized that increased

economic status of females would make it easier for females to exit an abusive relationship and

may possibly threaten males' control over their partners. Dugan et al. (1999) found that "net of

other changes, more educated women are better able, and perhaps more willing, to exit violent

relationships and thus avoid killing their partner" (205), however improved status appears to









homicide and whether or not similar theoretical factors influence male and female intimate

partner homicide.

Importance of Structural Changes and Domestic Violence Awareness

Declines in both male- and female-perpetrated intimate partner homicide rates were not

the only marked changes witnessed during the 1990 and 2000 time period. Changes were also

seen in domesticity, the status of women, levels of economic deprivation and marginalization,

and the availability of domestic violence resources. Any of these structural changes may have

had an impact on the observed patterns in intimate partner homicide.

From 1990 to 2000, there were significant changes in marital and non-marital

domesticity. Marriage rates plummeted and divorce rates increased, while the percent of

unmarried partner households increased from 3.5% in 1990 to 5.2% in 2000 (US Census Bureau,

2003). Declining marriage rates and increasing divorce rates influence intimate partner homicide

in terms of opportunity. With fewer people entering marriage and more getting divorced there is

less opportunity for violent married intimate partners to kill each other. However, while rates of

marriage and divorce are changing so are the rates of unmarried partners living together. With

more and more individuals choosing to live together and postpone marriage, the opportunity

increases for intimate partner homicide between non-married intimate partners.

Furthermore, the inherent dynamics of this type of relationship is different from a marital

relationship and may have an effect on the rate of intimate partner homicide.

Moreover, the status of females increased from 1990 to 2000. Females' employment

status, income, and educational attainment all improved. For instance, the percentage of females

16 and over that were employed in 1989 was 54.3%, compared with 57.4% in 1999 (Bureau of

Labor Statistics, 2008a). Improvements in the status of females would afford females the

opportunity to be able to leave an abusive relationship before the violence resulted in death to









partners. Unmarried partners include male householders living with a female partner and female

householders living with a male partner. This measure was calculated by dividing the number of

unmarried partner households by the total number of households and then multiplying this by

100 to obtain a percent.

(No. Unmarried Partner Households) x100
(Total no. Households)

Shelter rate. To measure the rate of the number of shelters available to females a measure

of the number of shelters for each city and year per 100,000 females 15 years and older was

computed. The number of shelters was divided by the total number of females aged 15 years and

older and then multiplied by 100,000 to obtain a rate. This is similar to Dugan et al. (1999)'s

calculation for the availability of hotlines, counseling, and legal services.

[ (Number of shelters for each city and year) x
x 100,000
(Number of females aged 15 years and older)

Legal service rate. The legal service rate is designed to measure the availability of legal

assistance, such as in obtaining restraining orders, court accompaniment, legal clinics or

advocacy. To measure the rate for the number of legal services available to females a measure of

the number of legal services for each city and year per 100,000 females 15 years and older was

computed. The number of legal services was divided by the total number of females aged 15

years and older and then multiplied by 100,000 to obtain a rate.

(Number of legal services for each city and year)x
x 100,000
(Number of females aged 15 years and older)

Number of referrals rate. Many cities may not have offered a particular service (i.e.,

shelters, hotlines, counseling, children's counseling, and legal services) but were able to refer

individuals to other programs that offered the service. To capture this, a measure for the rate of









work during the last 4 weeks, and were available to start ajob. Also included as unemployed

were civilians 16 years old and over who: did not work at all during the reference week, were on

temporary layoff from a job, had been informed that they would be recalled to work within the

next 6 months or had been given a date to return to work, and were available to return to work

during the reference week, except for temporary illness.

Civilian labor force Consists of people classified as employed or unemployed in accordance

with the criteria described above.

Labor force All people classified in the civilian labor force (i.e., "employed" and

"unemployed" people), plus members of the U.S. Armed Forces (people on active duty with the

United States Army, Air Force, Navy, Marine Corps, or Coast Guard).

Not in labor force All people 16 years old and over who are not classified as members of the

labor force. This category consists mainly of students, individuals taking care of home or family,

retired workers, seasonal workers enumerated in an off-season who were not looking for work,

institutionalized people (all institutionalized people are placed in this category regardless of any

work activities they may have done in the reference week), and people doing only incidental

unpaid family work (fewer than 15 hours during the reference week).

Household A household includes all of the people who occupy a housing unit. (People not

living in households are classified as living in group quarters.) A housing unit is a house, an

apartment, a mobile home, a group of rooms, or a single room occupied (or if vacant, intended

for occupancy) as separate living quarters

Householder The data on relationship to householder were derived from the question, "How is

this person related to Person 1," which was asked of Persons 2 and higher in housing units. One

person in each household is designated as the householder (Person 1). In most cases, the









G Correlation Matrices for Domestic Violence Service Variables. ............................118

R E F E R E N C E S ......... ..... ........................................................ .................................119

B IO G R A PH IC A L SK ETCH ........................................................................................ 129



















































6











Table 3-1. Means, standard deviations (in parentheses) for all variables, 1990 and 2000 (N= 178).

1990 2000


Female-perpetrated intimate partner homicide

(counts summed over 3 years)

Male-perpetrated intimate partner homicide
(counts summed over 3 years)

Percent female divorced

Percent male divorced

Percent unmarried households

Male batterers counseling per 100,000 males

Shelters per 100,000 females

Legal service programs per 100,000 females

Referral services per 100,000 females

Ratio of male to female education

Ratio of male to female income

Ratio of male to female employed

Female economic deprivation Index

Male economic deprivation index

Percent hispanic

Residential mobility

Officer rate per 1,000


6.553(10.864) 2.970(4.779)


11.569(18.378)


13.31(4.24)

9.92(3.12)

3.91(1.54)

.746(1.09)

.354(.276)

.313(.289)

.089(.243)

1.33(.140)

1.42(.173)

1.24(.106)

29.38(11.97)

26.97(11.28)

13.40(15.67)

50.22(7.88)

1.98(.910)


.303(.461)


8.770(13.929)


12.25(2.08)

9.31(2.03)

5.25(1.068)

.552(.738)

1.49(.909)

1.42(1.167)

.075(.187)

1.14(.098)

1.55(.180)

1.19(.086)

15.88(6.22)

14.88(5.88)

18.61(18.17)

51.40(6.71)

2.22(1.04)

.303(.461)


South









yit = xit + vi + + it i = 1,...,N; t =1,..., T

where xit is the vector of covariates for both cities (i) and time periods (t), with dummy variables

controlling for city (vi) and time trends (Xt). A time variable for year 1990 was included in the

analysis to control for period effects, with the year 2000 as the reference period. There are

advantages to this type of analysis. For instance, pooled cross-sectional time series analysis

increases the number of observations (i t) and also allows one to model time and space and

generalize between them.

Due to the skewed nature of the dependent variables both the female-perpetrated intimate

partner homicide count over 3 years and the male-perpetrated intimate partner homicide count

over 3 years was transformed using a logarithmic transformation. An issue arose with the

logarithmic transformation when the count of intimate partner homicide equaled zero. The log of

a zero equals -co. To fix this problem, all zero counts were also coded as zero for the logarithmic

transformation of the intimate partner homicide count. To account for this a dummy variable was

included in the models that represented zero counts for intimate partner homicide (See Hausman,

Hall, and Griliches, 1984 for rationale). Also, the log of the population 15 and older multiplied

by a factor of 3 was included in the models to control for the population size.

Both fixed-effects and random-effects models are appropriate to use for pooled cross-

sectional analysis, however for the current study the fixed effect estimation was used for a



meet the asymptotic assumptions of the Hausman test. The assumptions of the Hausman specification test are often

hard to meet. The Hausman test assumes that one of the estimators is efficient (i.e., has minimal asymptotic

variance). This assumption is violated if observations are clustered or pweighted or if the model is misspecified

(STATA 2008). For comparison purposes, both female-perpetrated and male perpetrated intimate partner homicide

were modeled using OLS fixed effects regression.









Specifically, the economic marginalization hypothesis suggests that as the economic

disadvantage of women increases, so will female crime. Research has found support for this

hypothesis in regards to female perpetrated offenses including intimate partner homicide (Box

and Hale, 1983; 1984; Steffensmeier and Streifel, 1992; Dugan et al., 2003; Reckdenwald and

Parker, 2008). To test the economic marginalization hypothesis, research has measured the

economic disadvantage of females in terms of female poverty and joblessness. For example, Box

and Hale (1983; 1984) found that female unemployment was positively associated with female

violent and property offenses. Importantly, research has shown that there is an overlap between

factors that effect male and female offending (Boritch and Hagan, 1990; Steffensmeier and

Haynie, 2000a; Steffensmeier and Haynie, 2000b). For instance, Steffensmeier and Haynie

(2000a) examined the effects of structural disadvantage (i.e., poverty, unemployment, income

inequality, female-headed households, and percent black) on male and female offending rates for

multiple offenses (i.e., homicide, robbery, aggravated assault, burglary, larceny-theft).

Steffensmeier and Haynie (2000) concluded that structural disadvantage variables influenced

female and male offending rates in similar ways. Surprisingly, the effects of the structural

variables were stronger for male offending.

The majority of studies have examined gender inequalities in relation to the killings of

females (Bailey and Peterson, 1995; Gauthier and Bankston, 1997; Avakeme, 1999), but

examining the impact of gender inequality on male homicides is important as well (Whaley and

Messner, 2002). For instance, Whaley and Messner (2002) explored the effects of both gender

inequality and economic deprivation on gendered-homicides (i.e., male offender-male victim,

male offender-female victim, female offender-female victim, and female offender-male victim)

in large cities for 1990 to 1994. Whaley and Messner measured gender inequality as the ratio of









CHAPTER 2
LITERATURE REVIEW

While research on intimate partner homicide is growing, the majority of research on

intimate partner homicide is directed at describing victimization and/or offending patterns (Block

and Christakos, 1995; Browne and Williams, 1993; Gallup-Black, 2005; Puzone et al., 2000;

Reidel and Best, 1998; Shackelford and Mouzos, 2005) rather than analyzing these patterns by

controlling for the impact of key factors that may have an influence on intimate partner homicide

over time (Block and Christakos, 1995; Browne and Williams, 1993; Puzone et al., 2000; Reidel

and Best, 1998). This research has found that the trends in intimate partner homicide during this

time period vary not only by the gender of the victim and offender, but also by the type of

relationship between the victim and offender.

For instance, Puzone et al. (2000) examined U.S. homicide victimization from 1976-1995

and found that the decline in homicide victimization rates were more pronounced for husbands

(75%) than wives (41%), boyfriends (68%) more than girlfriends (21%)6, and ex-husbands

(88%) more than ex-wives (41%). Specifically, Puzone et al. (2000) found that the risk of being

killed by a spouse in 1995 was 2.6 times greater for a wife than a husband, 2 times greater for a

girlfriend than a boyfriend, and for the combined period of 1992-1995 was 3.8 times greater for

an ex-wife than an ex-husband. Earlier work examining male- and female-perpetrated intimate

partner homicide supports these findings (Browne and Williams, 1993). That is, Browne and

Williams (1993) examined Supplemental Homicide Report data from 1976-1987 for all states

and found that both married male-perpetrated (i.e., males killing their female spouses) and

female-perpetrated (i.e., females killing their male spouses) intimate partner homicide declined,

however the decline was more pronounced for the rate of wives killing husbands than the rate of


6 However, this decline was not statistically significant.









100,000 females is associated with an increase in female-perpetrated intimate partner homicide.

One reason for this finding may be that although the shelter rate is accounting for the number of

shelters available per 100,000 females in cities, it may not be capturing shelter usage by females

in abusive relationships. It is possible that the availability of shelters across cities is increasing

but females are not taking advantage of them to reduce the exposure between themselves and

their abusive partner. On the opposite end of the spectrum, females may in fact be using shelters

to reduce the exposure between themselves and their violent partners, but once shelter stays end

may find themselves in contact with their violent partners and may have to resort to lethal

violence to protect themselves, if adequate protection is not available. Furthermore, research has

shown that the most dangerous time in an abusive relationship is when a partner is trying to end

the relationship (Goetting, 1995). Some females may be reacting to male partners' violence over

the threat of the end of the relationship and may resort to lethal means to protect themselves.

The second domestic violence service variable associated with female-perpetrated

intimate partner homicide is the number of legal services per 100,000 females. This variable is

associated with a decrease in female-perpetrated intimate partner homicide. It seems that as

females are provided with more opportunities for legal assistance, such as in obtaining

restraining orders, court accompaniment, legal clinics or advocacy the opportunities for females

to kill their male partners decreases.

Interestingly, three of the six exposure reduction measures are associated with male-

perpetrated intimate partner homicides in 1990 and 2000. Similar to female-perpetrated intimate

partner homicide in 2000, the percent of males divorced is positively related to male-perpetrated

intimate partner homicide. Based on suggestions from prior research (Goetting, 1995), it seems









killed by an intimate has decreased during this same time period (5.3% and 2.8%, respectively).

This is evident in Figure 1-3. Specifically, from 1989 to 2001 there has been a 13% increase in

the proportion of females killed by an intimate partner compared to a 47% decrease in the

proportion of males killed by an intimate partner. In addition, the intimate partner is the only

victim-offender relationship category where female-perpetrated homicide rates approach that of

men (Browne, Williams, and Dutton, 1999). This is apparent when examining the difference

(i.e., the gap) between male- and female-perpetrated intimate partner homicide rates.

Typically, males commit roughly 90% of all homicides occurring in the United States,

making males 10 times more likely than females to commit murder (Bureau of Justice Statistics,

2006). Males are more likely than females to be perpetrators for every victim-offender

relationship category (i.e., intimate partners, family members, acquaintances, and strangers).

However, when females kill they are more likely to kill someone that is close to them, such as an

intimate partner or a family member, making the intimate partner the only victim-offender

relationship where female-perpetration approaches that of men (Browne, Williams, and Dutton,

1999).

The gap in the difference between male- and female-perpetrated intimate partner

homicide rates for large cities has narrowed in recent years. This is apparent when examining

Figure 1-2. The gap between the rates of male- and female-perpetrated intimate partner homicide

was fairly stable from 1989 until 1992. In 1992, the rate of intimate partner homicide increased

for both males and females, but far more drastically for males killing their female-partners. This

difference in the rise in the rates of intimate partner perpetration resulted in a widening of the

gap between the rates of male- and female-perpetrated intimate partner homicide. As the rates of

male- and female-perpetrated intimate partner homicide plummeted and then leveled off after









that lower marriage rates are related to fewer husbands being killed by their wives. Rosenfeld

(1997) suggests that "much of the decline in intimate partner homicide is a function of change in

the rate of marriage with the age groups at highest risk of homicide victimization and offending"

(73).

However, decreasing marriage rates may mean that more individuals are cohabitating

together without getting married. Cohabitation has been shown to be an important risk factor in

intimate partner homicide (Shackelford and Mouzos, 2005). Research suggests that males and

females in cohabitating relationships are at a higher risk of intimate partner homicide

victimization compared to males and females in married relationships (Daly and Wilson, 1988;

Shackelford, 2001; Wilson, Daly, and Wright, 1993; Wilson, Johnson, and Daly, 1995). For

instance, Wilson, Daly and Wright (1993) and Wilson, Johnson and Daly (1995) found that

females that cohabitate with their partner are 9 times more likely to be killed by their intimate

partner than are married females. In addition, Shackelford (2001) found that men that are

cohabitating with their female partners are 10 more likely to be victims of intimate partner

homicide compared to men in married relationships. Rodriquez and Henderson (1995) suggest

that "individuals in cohabitating relationships, having invested heavily in such relationships, may

be more likely to resort to violent retaliation (i.e., homicide) than those in dating or non-

cohabitating relationships because they do not perceive themselves as having legal 'protection'

associated with marriage" (48).

Improved economic status of women.

Improved economic status of women has also been considered to produce exposure

reduction effects. Improved economic status of women in terms of educational attainment,

income, and employment increases the access of women to opportunities that may give them









1994, the gap in perpetration rates began to consistently narrow. The rate of male-perpetrated

intimate partner homicide gradually decreased while the rate of female-perpetrated intimate

partner homicide remained fairly stable, resulting in a gradual narrowing of the gap in

perpetration rates that continued through 2001.

Figure 1-4 summarizes the decline in the rate of total U.S. homicides in large cities, as

well as the decline in the rates of male- and female-perpetrated intimate partner homicide from

1990 to 2000. The total rate of homicide decreased 42% during this time period. Both male- and

female-perpetrated intimate partner homicides decreased as well. The rate of female-perpetrated

intimate partner homicide had the largest decrease of 52% compared to 34% for male-

perpetrated intimate partner homicide. No confirmed explanation is available for the greater

intimate partner homicide rates among males compared to females or the gender-specific

differences that have been observed over time.

Overview of Theoretical Perspectives

Three theoretical perspectives are of interest in this study the exposure reduction

hypothesis, the backlash hypothesis, and economic deprivation and marginalization. Previous

research has supported each of these perspectives; therefore all three perspectives are believed to

influence intimate partner homicide uniquely.

The exposure reduction hypothesis proposes that the reduction in the time spent (i.e.,

exposure) between couples in violent relationships results in less intimate partner homicide,

because couples have less contact with one another and therefore have less opportunities to kill

their intimate partner. A number of factors have been considered to cause exposure reduction

effects on intimate partner homicide, such as declining marriage rates, increasing divorce rates,

greater economic and educational status of females, and the availability of domestic violence

resources and services (Dugan et al., 1999; 2003; Rosenfeld, 1997).









REFERENCES


Agresti, Allen. 1996. An Introduction to Categorical Data Analysis. New York: John Wiley and
Sons.

Allard, Mary Ann, Randy Albelda, Mary Ellen Colten, and Carol Cosenza. 1997. In Harms
Way? Domestic Violence, AFDC Receipt, and Welfare Reform in Massachusetts.
Boston, MA: University of Massachusetts.

Allison, Paul D. 1994. Using panel data to estimate the effects of events. Sociological Methods
and Research 23: 179-199.

Anderson, Kristin L. 1997. Gender, status, and domestic violence: An integration of feminist and
family violence approaches. Journal ofMarriage and the Family 59:655-669.

Archer, John. 2000. Sex differences in aggression between heterosexual partners: A Meta-
analytic review. Psychological Bulletin 126:651-680.

Avakeme, Edem F. 1999. Females' labor force participation and intimate femicide: An empirical
assessment of the backlash hypothesis. Violence and Victims 14:277-291.

Bailey, William C. 1984. Poverty, inequality, and city homicide rates: Some not so unexpected
results. Criminology 22:531-550.

Bailey, William C., and Ruth D. Peterson. 1995. Gender inequality and violence against women:
The case of murder. In Crime and Inequality, eds. John Hagan and Ruth D. Peterson.
Stanford, CA: Stanford University Press.

Baron, Larry, and Murray A. Straus. 1987. Legitimate violence, violent Attitudes, and rape: A
test of the cultural spillover theory. Annals of the New York Academy of Sciences
528:79-110.

Berk, Richard A., Sarah F. Berk, Donileen R., Loseke, and David Rauma. 1983. Mutual combat
and other family violence. In The dark side offamilies: Current family violence
research, eds. David Finkelhor, Richard J. Gelles, Gerald T. Hotaling, & Murray A.
Staus. Beverly Hills, CA: Sage.

Barnard, George W, Hernan Vera, Maria I. Vera, and Gustave Newman. 1982. Till death do us
part: A study of spouse murder. Bulletin of the American Academy of Psychiatry and
the Law 10:271-280.

Blau, Judith R., and Peter M. Blau. 1982. The cost of inequality: Metropolitan structure and
violent crime. American Sociological Review 47:114-129.

Block, Carolyn R., and Anigone Christakos. 1995. Intimate partner homicide in Chicago over 29
Years. Crime & Delinquency 41:496-526.









BIOGRAPHICAL SKETCH

Amy Reckdenwald grew up in Ithaca, New York. After graduating from high school she

moved to Pittsburgh, Pennsylvania where she obtained her bachelor's degree in psychology with

a double major in statistics from Carnegie Mellon University. After completing her

undergraduate degree in 2001 she moved to Tampa, Florida and later received her master's

degree in criminology from the University of South Florida in 2004. In 2008 she completed her

Ph.D. in criminology, law and society at the University of Florida in Gainesville, Florida. After

graduating from the University of Florida, Dr. Amy Reckdenwald decided to remain in Florida

and accepted a position as an Assistant Professor at Florida Atlantic University, in Boca Raton,

Florida.




Full Text

PAGE 1

1 EXAMINING CHANGES IN MALE AND FE MALE INTIMATE PARTNER HOMICIDE OVER TIME, 1990-2000 By AMY RECKDENWALD A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2008

PAGE 2

2 2008 Amy Reckdenwald

PAGE 3

3 ACKNOWLEDGMENTS First and forem ost, I would like to thank all of my family and friends for all their love and encouragement through my many year s of college. I honestly would never have finished without their endless support and I know they are as happ y as I am to see my graduate school career finally come to an end. I would also like to thank my entire dissertation committee, Dr. Karen Parker, Dr. Lonn Lanza Kaduce, Dr. Alex Piquero, and Dr. Barbara Zsembik, for all their support and suggestions in the writing of this diss ertation. I especially want to thank Dr. Karen Parker for all her help and guidance through my 4 years at UF as well as all the assistance she gave in the completion of this dissertation. She is a mentor as well as a good friend. I would also like to thank Dr. Richard Hollinger for sitting in on my defense last mi nute and giving me many good suggestions to improve this dissertation a nd Dr. Larry Winner for all his help with the statistical analysis.

PAGE 4

4 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................3LIST OF TABLES................................................................................................................. ..........7LIST OF FIGURES.........................................................................................................................8ABSTRACT.....................................................................................................................................9CHAPTER1 INTRODUCTION..................................................................................................................11Decline in Intimate Partner Homicide.................................................................................... 11Importance of Examining Intimate Partner Homicide............................................................ 12Overview of Theoretical Perspectives.................................................................................... 14Importance of Structural Changes and Domestic Violence Awareness................................. 17Research Questions............................................................................................................. ....20Significance of Study..............................................................................................................20Summary.................................................................................................................................222 LITERATURE REVIEW.......................................................................................................28Theoretical Perspectives....................................................................................................... ..30Exposure Reduction Hypothesis...................................................................................... 30Factors that Reduce Exposure......................................................................................... 32Exposure Reduction Predictions...................................................................................... 37Backlash or Retaliation Hypothesis................................................................................. 38Backlash Predictions.......................................................................................................41Economic Deprivation and Marginalization................................................................... 41Economic Deprivation and Marginalization Predictions................................................ 45Summary..........................................................................................................................45Hypotheses..............................................................................................................................45Exposure Reduction......................................................................................................... 46Backlash..........................................................................................................................46Economic Deprivation and Marginalization................................................................... 47Change Models................................................................................................................ 473 DATA AND METHODOLOGY........................................................................................... 49Unit of Analysis............................................................................................................... .......49Sources of Data.......................................................................................................................50Homicide Data.................................................................................................................. ......50Domestic Violence Service Data............................................................................................ 51Measures.................................................................................................................................52

PAGE 5

5 Dependent Variable......................................................................................................... 52Independent Variables.....................................................................................................52Exposure reduction measures................................................................................... 53Backlash measures................................................................................................... 55Economic deprivation and marginalization measures.............................................. 56Controls.................................................................................................................... 58Missing Data...........................................................................................................................59Multicollinearity.....................................................................................................................61Methodology...........................................................................................................................63Descriptive Statistics....................................................................................................... 63Analytical Plan................................................................................................................ ........65Cross-Sectional Analyses................................................................................................65Pooled Cross-Sectional Time Series Analysis: Fixed Effect Estimation........................ 664 RESULTS...............................................................................................................................74Cross-Sectional Analysis........................................................................................................74Female-Perpetrated Intimate Partner Homicide 1990..................................................... 74Male-Perpetrated Intimate Partner Homicide 1990......................................................... 76Female-Perpetrated Intimate Partner Homicide 2000..................................................... 77Male-Perpetrated Intimate Partner Homicide 2000......................................................... 79Pooled Cross-Sectional Time Series Analysis: Fixed Effect Estimation........................ 815 DISCUSSION AND CONCLUSIONS.................................................................................. 90Discussion...............................................................................................................................90Exposure Reduction......................................................................................................... 91Backlash..........................................................................................................................93Economic Deprivation and Marginalization................................................................... 95Summary..........................................................................................................................95Changes Over Time.........................................................................................................97Limitations.................................................................................................................... ..........98Future Research......................................................................................................................99Conclusion............................................................................................................................100APPENDIX A Glossary................................................................................................................................103B Correlation Matrix for Explanat ory and Outcome Variable 1990........................................ 109C Correlation Matrix for Explanat ory and Outcome Variables 2000...................................... 111D Simplified Correlation Matrix for Explanatory and Outcome Variables 1990....................113E Simplified Correlation Matrix for Expl anatory and Outcome Variables 2000....................115F Variance Inflation Factors for Simplified Models................................................................ 117

PAGE 6

6 G Correlation Matrices for Domes tic Violence Service Variables. ......................................... 118REFERENCES............................................................................................................................119BIOGRAPHICAL SKETCH.......................................................................................................129

PAGE 7

7 LIST OF TABLES Table page 3-1Means, standard deviations (in parentheses) for all variables, 1990 and 2000 (N= 178). 693-2T-test scores for change from 1990 to 2000 (N=178)....................................................... 703-3Variance inflation factors for all variables included in the models (N=178).................... 713-4Principal components analysis af ter varimax rotation 1990 (N=178)............................... 724-1Zero-inflated negative binomial regression equations with coefficients (and ZScores) for gender-specific in timate partner homicide 1990............................................. 844-2Zero-inflated negative binomial regression equations with coefficients (and ZScores) for gender-specific in timate partner homicide 2000............................................. 854-3Summary of cross-sectiona l results, 1990 and 2000..........................................................864-4Fixed-effects regression coefficients fo r the relationship between changes in predictor variables and cha nges in logged female intimate partner homicide counts for 143a US cities, 1990-2000............................................................................................ 874-5Fixed-effects regression coefficients fo r the relationship between changes in predictor variables and cha nges in logged male intimate partner homicide counts for 174a US cities, 1990-2000.................................................................................................884-6Summary of change model results, 1990 to 2000.............................................................. 89

PAGE 8

8 LIST OF FIGURES Figure page 1-1Total homicide rate per 100,000 reside nts for large cities, 1989-2001............................. 241-2Gender-specific intimate partner homic ide rates for larg e cities, 1989-2001....................251-3 Proportion of all homicides invol ving intimate partners, 1989-2001............................... 261-4Summary of percent change from 1990 to 2000 in total homicide rate, femaleperpetrated intimate partner homicide rate and male-perpetrated intimate partner homicide rate.................................................................................................................. ....272-1Detailed theoretical and conceptual model on exposure reduction, backlash, economic deprivation and marg inalization and maleand fe male-perpetrated intimate partner homicide............................................................................................................... .48

PAGE 9

9 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy EXAMINING CHANGES IN MALE AND FE MALE INTIMATE PARTNER HOMICIDE OVER TIME, 1990-2000 By Amy Reckdenwald August 2008 Chair: Lonn Lanza-Kaduce Cochair: Karen F. Parker Major: Criminology, Law and Society Research on intimate partner homicide has increas ed in the recent years. This is partially due to the dramatic decline witn essed over the last couple of decad es in these types of homicides as well as the growth that has occurred in public awarene ss and policy responses toward domestic violence. Recent intimate partner hom icide research has predominately focused around two perspectives to explain th e relationship between intimate partner homicide and domestic violence resources the exposure reduction hyp othesis and the backlash or retaliation hypothesis, with results that support both (Dugan, Nagin, and Rosenfeld, 1999; 2003). The exposure reduction hypothesis proposes that do mestic violence resources that reduce the exposure or contact between intimate partners shou ld decrease the probability of intimate partner homicide, while the backlash hypothesis suggests the opposite. That is, domestic violence interventions may have unintended consequences and increase the risk of intimate partner homicide if they threaten male domin ance and control over their partners. The contradictory results from these studies have fr ustrated advocates and have made them question their efforts to make females safer. Societal remedies to lowe r intimate partner homicide seem to be addressing male-perpetrated intimate partner violence and homi cide insufficiently. It is important to gain a

PAGE 10

10 better understanding of what fact ors are truly influencing gende r-specific intimate partner homicide, during a time that is marked by signi ficant transformations in domestic violence legislation and response toward domestic violence with the enactment of the Violence Against Womens Act of 1994. This research helps to address the contradi ctory findings while controlling for a number of structur al factors that have shown to be important in the homicide literature. The current study examines both the ar guments of exposure reduction and backlash, in addition to economic deprivation an d marginalization to explain th e observed patterns in maleand female-perpetrated intimate partner homicide over time, something that has not been done in the literature to date. Measures of key concepts will be co llected for 2 decennial points (1990 and 2000). Supplemental homicide files for this time pe riod and census data will be utilized. Poisson regression models will be used to investigate wh ich theoretical perspective is associated with maleand female-perpetrated intimate part ner homicide in 1990 and 2000. Pooled crosssectional fixed effect time series regression will be used to determine whether changes from 1990 to 2000 in key structural indicators influe nce trends in maleand female-perpetrated intimate partner homicide during this same time period. Overall, resu lts suggest that key theoretical indicators do in fact influence the trends in female-perpetrated intimate partner homicide, but not male-perpetrated intimate partner homicide.

PAGE 11

11 CHAPTER 1 INTRODUCTION Only in the area of partner hom icides do wo mens perpetration rates approach that of men. (Browne, Williams, and Dutton 1999: 150) Decline in Intimate Partner Homicide Research on intim ate partner homicide1 has increased in recent years. Increased attention toward intimate partner homicide has resulted from the dramatic decline in perpetration that has been witnessed in these types of homicides. This decline coincided with the remarkable decline that was observed in total homicides rates that still puzzles researchers. Beginning in the early 1990s and extending into the 21st century, Amer icans witnessed an unprecedented drop in violent crime rates, specifically a drop in hom icide rates. This drop was witnessed across the U.S., but was seen in large cities2 in particular. From 1990 to 2000, ho micide rates in large cities dropped roughly 42%3. See Figure 1-1 for an illustration of the decline in total U.S. homicide rates in large cities4. Interestingly, the decline in adult homicide rates ha s been recorded for all victim-offender relationship categor ies, but the sharpest decline in homicides have been observed in the family category for both males and fe males (Rosenfeld, 2000), with most of those homicides generally occurring between spouses, ex-spouses, boyfriends, and girlfriends (Bureau of Justice Statistics, 200 6; Durose et al., 2005). Statistics reveal that U.S. intimate partner homicide c ounts have been declining for over two decades (Bureau of Justice Statistics, 2006 ; Greenfield et al., 1998). See Figure 1-2 for 1 Intimate partners include spouses, ex-spouses, or boyfriends and girlfriends. 2 Large cities have a population of 100,000 residents or more. 3 See Figure 1-4 for summary of percent change from 1990 to 2000. 4 Homicide rates per 100,000 population were calculated by dividing the homicide counts by the population and then multiplying that number by 100,000 to obtain a rate. Denominators were obtained from the 1990 and 2000 U.S. Census Bureau. Annual population estimates were calcula ted for between census years by subtracting the 1990 population from the 2000 population and then dividing that difference over time by 10 (i.e., number of years). The annual population estimate was then subtracted from the 1990 population to obtain a population estimate for 1989 and added to 1990 and each year after to obtain a population estimate for each of the following years.

PAGE 12

12 gender-specific total intimate partner ho micide rates for large cities, 1989-20015. Surprisingly, maleand female-perpetrated intimate partner homicide rates appear to follow a similar pattern, though rates of male-perpetrated intimate partner homicide are higher than female-perpetrated rates. That is, both male and female intimate part ner homicide rates fluctuat e but are fairly stable from 1989 until 1992. From 1992 to 1994 a dramatic spike in the rates of intimate partner homicide is apparent. After 1994, intimate partner homicide rates level off. Female-perpetrated intimate partner homicide rates remain fairly stable from 1995 to 2001, while male-perpetrated intimate partner homicide rates decline slightl y. Though, the rates of male and female intimate partner homicide show a similar pa ttern over time, a greater decline is observed in the rate of female-perpetrated intimate partner homicide. Speci fically, the rate of male-perpetrated intimate partner homicide decreased 33% from 1989 to 2001, whereas the rate of female-perpetrated intimate partner homicide decreased 45% during this same time. The reason for the more substantial decrease in the rate of female-perpe trated intimate partner homicide compared with male-perpetrated intimate partner homicide is unknown and remains an interest to researchers. Importance of Examining In timate Partner Homicide There are many reasons that make intimate partner homicide an important topic for research. First and foremost, as stated earlier, in timate partner homicide ha s decreased in recent years with no concrete explanation. Furthermor e, females are more likely to be killed by an intimate partner than any other offender. Among female homicide victims, roughly 1 in 3 is killed by an intimate partner, whereas among male homicide victims, only 5% are killed by an intimate partner. Moreover, although intimate partner homicide rates have been decreasing, the proportion of female homicide victims killed by an intimate partner has gradually increased from 1989 to 2001 (27.7% and 31.4%, respectively), whereas the proportion of male homicide victims 5 See footnote 4 for the rate calculation.

PAGE 13

13 killed by an intimate has decrea sed during this same time peri od (5.3% and 2.8%, respectively). This is evident in Figure 1-3. Specifically, fr om 1989 to 2001 there has been a 13% increase in the proportion of females killed by an intimate partner compared to a 47% decrease in the proportion of males killed by an intimate partner. In addition, the intimate partner is the only victim-offender relationship category where female-p erpetrated homicide ra tes approach that of men (Browne, Williams, and Dutton, 1999). This is apparent when examining the difference (i.e., the gap) between maleand female-perpe trated intimate partner homicide rates. Typically, males commit roughly 90% of all homicides occurring in the United States, making males 10 times more likely than females to commit murder (Bureau of Justice Statistics, 2006). Males are more likely than females to be perpetrators for every victim-offender relationship category (i.e., intimat e partners, family members, acquaintances, and strangers). However, when females kill they are more likely to kill someone that is close to them, such as an intimate partner or a family member, making th e intimate partner the only victim-offender relationship where female-perpetration approach es that of men (Browne, Williams, and Dutton, 1999). The gap in the difference between malea nd female-perpetrated intimate partner homicide rates for large cities has narrowed in recent years. This is apparent when examining Figure 1-2. The gap between the ra tes of maleand female-perpetrated intimate partner homicide was fairly stable from 1989 until 1992. In 1992, the rate of intimate partner homicide increased for both males and females, but far more drastically for males killing their female-partners. This difference in the rise in the rates of intimate partner perpet ration resulted in a widening of the gap between the rates of malea nd female-perpetrated intimate partner homicide. As the rates of maleand female-perpetrated intimate partner ho micide plummeted and then leveled off after

PAGE 14

14 1994, the gap in perpetration rates began to consistently narrow. The rate of male-perpetrated intimate partner homicide gradually decreased while the rate of female-perpetrated intimate partner homicide remained fairly stable, resu lting in a gradual narrowing of the gap in perpetration rates that continued through 2001. Figure 1-4 summarizes the decline in the rate of total U.S. homicides in large cities, as well as the decline in the rates of maleand female-perpetrated intimate partner homicide from 1990 to 2000. The total rate of homicide decreased 42% during this time period. Both maleand female-perpetrated intimate partner homicides decr eased as well. The rate of female-perpetrated intimate partner homicide had the largest de crease of 52% compared to 34% for maleperpetrated intimate partner homicide. No confir med explanation is available for the greater intimate partner homicide rates among males compared to females or the gender-specific differences that have been observed over time. Overview of Theoretical Perspectives Three theoretical perspectives are of interest in this study the exposure reduction hypothesis, the backlash hypothesis, and economic deprivation and m arginalization. Previous research has supported each of th ese perspectives; theref ore all three perspec tives are believed to influence intimate partner homicide uniquely. The exposure reduction hypothesis proposes that the reduction in the time spent (i.e., exposure) between couples in violent relationshi ps results in less intimate partner homicide, because couples have less contact with one another and therefore have less opportunities to kill their intimate partner. A number of factors have been consid ered to cause exposure reduction effects on intimate partner homicide, such as decl ining marriage rates, increasing divorce rates, greater economic and educational status of female s, and the availability of domestic violence resources and services (Dugan et al., 1999; 2003; Rosenfeld, 1997).

PAGE 15

15 The backlash hypothesis proposes that the incr eased status of females threatens males control and dominance over females and will resu lt in more intimate partner homicide, because males will use more violence to dominate and cont rol their female partners. Factors that have been thought to produce backlash ef fects overlap with some of the factors that are considered to produce exposure reduction effects. Some of the factors considered include: greater economic, employment, and educational status of females and the availability of domes tic violence services when they fail to adequately reduce contact betw een violent partners (D ugan et al., 1999; 2003). Research supports these two main theoretical perspectives to explain intimate partner homicide (Dugan et al., 1999; 2003). However, the contradictory results from these studies have frustrated advocates. For instance, Dugan et al. (1999) reported an exposur e reduction effect of domestic violence resources on male intimate part ner homicide victimization, but not on female intimate partner homicide victimization, suggesti ng that the movement to prevent domestic violence only really made men safer and not women. In addition, Dugan et al. (2003) reported that some domestic violence services actually crea te a backlash effect and seem to put some women in more danger of intimate partner homic ide victimization (i.e., prosecutors willingness to prosecute). Both of these perspectives need to be examined more closely to determine which factors are truly influencing gender-s pecific intimate partner homicide. Due to the contradictory findings from the above two perspectives it is important to also consider other factors. This study draws from the economic deprivati on and marginalization literature to gain a better unders tanding of intimate partner homicide. Economic deprivation and marginalization have been shown to explain hom icide in general and th erefore seems reasonable to include in the present rese arch. Ideas behind economic depriv ation and marginalization come from both the strain and feminist literatures. Stra in theorists relate the opportunity structure to

PAGE 16

16 violence. Merton, ([1949] 1968: 185248) argues that deviance is a result of an unequal distribution of opportunities among classes in society. According to Merton, the distribution of opportunities in the social structur e determines which classes are most likely to be involved in crime and deviance. Males or females that lack certain opportunities becaus e of deprivation (i.e., poverty and unemployment) may be frustrated or strained and this strain may influence intimate partner homicide perpetration. On the other hand, feminists scholars tend to focus on inequalities between men and women. Feminist literature ex amining inequalities tend to concentrate on the economic marginalization hypothesis, which fo cuses on the economic disadvantage of women relative to men (Heimer, 2000). The economic marginalization hypothesis argues that the economic disadvantage of women (i.e., female unemployment and poverty) is a significant predictor of female crime (Box and Hale 1983, 1984; Heimer, 2000; Heimer, Wittrock, and Unal, 2005; Hunnicutt and Broidy, 2004; Steffensmeier and Streifel 1992). Extant homicide research has incorporated the ideas behind ec onomic deprivation and ma rginalization (Bailey, 1984; Blau and Blau, 1982; Land, McCall, and Cohen, 1990; Parker, McCall, and Land, 1999; Steffensmeier and Haynie, 2000a; 2000b; Whaley and Messner, 2002); however limited research has included indicators of economic deprivation and marginalization to explain intimate partner homicide (for an exception see Reckdenwald and Parker, 2008). Economic deprivation and marginalization in terms of poverty or unempl oyment may influence male and female intimate partner homicide perpetration. All three theoretical perspectives have rece ived support in the liter ature; though at times some of the findings are mixed and even cont radictory. This study hopes to determine which theoretical factors significantl y contribute to maleand female-perpetrated intimate partner

PAGE 17

17 homicide and whether or not similar theoretica l factors influence male and female intimate partner homicide. Importance of Structural Changes and Domestic Violence Aw areness Declines in both maleand female-perpetrat ed intimate partner homicide rates were not the only marked changes witnessed during the 1990 and 2000 time period. Changes were also seen in domesticity, the status of women, leve ls of economic deprivation and marginalization, and the availability of domestic violence resource s. Any of these structural changes may have had an impact on the observed pattern s in intimate partner homicide. From 1990 to 2000, there were significant changes in marital and non-marital domesticity. Marriage rates plummeted and divorce rates increased, while the percent of unmarried partner households increased from 3.5% in 1990 to 5.2% in 2000 (US Census Bureau, 2003). Declining marriage rates a nd increasing divorce rates influe nce intimate partner homicide in terms of opportunity. With fewer people entering marriage and more getting divorced there is less opportunity for violent married intimate partners to kill each other. However, while rates of marriage and divorce are changing so are the rates of unmarried partners living together. With more and more individuals choosing to live t ogether and postpone marriage, the opportunity increases for intimate partner homicide between non-married intimate partners. Furthermore, the inherent dynamics of this type of relationship is different from a marital relationship and may have an effect on th e rate of intimate partner homicide. Moreover, the status of females increas ed from 1990 to 2000. Females employment status, income, and educational attainment all imp roved. For instance, the percentage of females 16 and over that were employed in 1989 was 54.3% compared with 57.4% in 1999 (Bureau of Labor Statistics, 2008a). Improvements in the st atus of females would afford females the opportunity to be able to leave an abusive relations hip before the violence resulted in death to

PAGE 18

18 themselves or their partners; thus, resulting in a lowering of intimate partner homicide. However, these improvements in females status may ha ve unintended consequences on intimate partner homicide. Some males may feel threatened by an in crease in the status of females and may react in a violent and possibly lethal ma nner towards their female partne rs to gain back the perceived loss of control. In addition to changing domesticity and impr oving status of females, the 1990s were a time of economic prosperity. The percent of indi viduals that reported liv ing in poverty decreased from 13.1% in 1989 to 12.4% in 1999 (U.S. Bureau of the Census, 2003). In addition to this, the unemployment rate decreased as well. In 1989, the percentage of the population that was unemployed was 5.3%, compared with 3.7% in 1999 (Bureau of Labor Statistics, 2008b). Economic deprivation may influence the dyna mics within the household. Less stress and frustration over lack of employment and living in poverty may result in a decrease in violence towards intimate partners. Domestic violence service and resource availabi lity increased in the 1970s, but more so after the enactment of the 1994 VAWA. During this time, domestic violence policies and programs that offered services for females and even males expanded considerably. The availability of domestic violence resources and services has been considered a key factor in the explanation of the decline in intimate partner homic ide (Dugan et al., 1999; 2003). Before the 1990s, many jurisdictions did not allow police to arrest on a misdemeanor charge unless they had witnessed the act. The battered Wo mens Movement pressed for redefining the act of domestic violence from a domestic disturbance (i.e., someth ing that is seen as a private family matter) to defining it as unacceptable and a criminal offense that will not be tolerated (Schechter, 1982). By the mid 1970s resources began to become availa ble to abused women. However, it could be

PAGE 19

19 argued that domestic violence against women was not seen as a national social problem until the enactment of the 1994 VAWA. The VAWA si gnificantly increased funding for domestic violence services, and supported domestic viol ence specialization in local police departments and prosecutor offices (Dugan et al., 2003: 171) Some examples of the VAWAs programs include policies to encourage the prosecution of abusers, victim services, and domestic violence prevention programs. Although this act specifi cally acknowledges violence against women, violence, both lethal and non-lethal, shoul d be acknowledged against both men and women. Domestic violence services should influen ce intimate partner homicide by limiting the opportunity for contact between violent intimate partners and thus limit the opportunities for partners to kill each other. Despite the reports that the increased availa bility of domestic violence services and resources is a significant factor in the explanation of the decline in intimate partner homicide over the past couple of decades (Browne and Williams, 1989; Dugan et al., 1999; 2003), resources have historically been directed at females and more so married females (Browne and Williams, 1993; Browne, Williams, and Dutton, 1999). This gives females options besides killing their partners and may in fact be a main reason that male intimate partner homicide victimization has decreased signi ficantly. However, this explan ation does not hold for female intimate partner homicide victimization. That is, female intimate partner homicide victimization has not seen as drastic of a reduction as ma le intimate partner homicide victimization. Taken together, there have been many change s witnessed over the la st couple of decades in intimate partner homicide as well as in domesticity, the status of women, economic deprivation, and in the availability of domestic violence resources. It is important to determine if these changes can account for the significant de cline in intimate partner homicide. Examination

PAGE 20

20 reveals that rates of female-perpetrated in timate partner homicide from 1990 to 2000 had a greater decline than male-perpetrated intimate pa rtner homicide. It is important to determine what factors account for this de cline as well as possible reasons why the decline was not as drastic for male-perpetrated intimate partner homicide. Research Questions One of the m ain purposes of this study is to examine gender-specific intimate partner homicide and determine which structural correlate s contribute to the observed patterns seen in 1990 and 2000. In addition, this study will examin e the changes between covariates in 1990 and 2000 to determine if any of the changes can ac count for the changes in maleand femaleperpetrated intimate partner homicide over time. The following research ques tions will be addressed: 1. What is the impact of structural measures on maleand female-perpetrated intimate partner homicide in 1990 and 2000? 2. Will criminological theories concerning domestic violence help explain maleand female-perpetrated intimate partne r homicide in 1990 and in 2000? 3. Does the availability of domestic violence services influence intimate partner homicide in 1990 and 2000? 4. What is the impact of change over time (1990-2000) in key stru ctural variables on maleand female-perpetrated intimate partner homicide trends? Significance of Study This study expands on the intimate partne r homicide literature in several ways. First, this study draws from three theoretical persp ectives in an effort to explain intimate partner homicide over time. The domestic violence lit erature has acknowledged both the importance of reducing the exposure between violent partners as well as backlash or retaliation by male

PAGE 21

21 partners as a response or cons equence to interventions designe d to reduce violence between couples. However, this research has found resu lts that support both pe rspectives. This study draws from two domestic violence perspectives while also controlling for a third theory of economic deprivation and marginal ization in attempt to more clearly understa nd trends in intimate partner homicide. This study examines both maleand femaleperpetrated intimate pa rtner homicide and compares their trends over time. This study not only documents change but also explains changes in intimate partner homicide trends over time and gives a better understanding of what is correlated with this observed change. Examining changes in key indicators on intimate partner homicide over time has not been done in the li terature. The dynamic nature of key indicators will be examined to determine whether or not th e changes from 1990 to 2000 in key indicators correlates with the change in maleand female -perpetrated intimate partner homicide over time. For example, it is of interest to determine if the change in the percent divorced in cities from 1990 to 2000 is related to the changes in intimate partner homicide during this same time period. Focusing on changes over time will give a better understanding of the differential trends in maleand female-perpetrated intimate partner homicide. The time period being examined in this study depicts a time where there has been significant changes in homicide itself as well as domesticity (i.e., living arrangements), the status of females, and economic deprivation and margin alization. Moreover, during this time, domestic violence was being seen as a national social problem and funding for domestic violence services were increasing significantly with the enactme nt of the 1994 Violence Against Womens Act (VAWA). This study controls for the influence of various domestic violence services while

PAGE 22

22 examining other structural factors in an effort to better understand maleand female-perpetrated intimate partner homicide over time. Summary This is a m acro-level study that proposes to build on existing literature by further considering the effects of key indicators that represent the exposure reduction and backlash perspectives, while also includi ng a third theoretical perspectiv e of economic deprivation and marginalization. Data will be collect ed to capture these theoretical perspectives in an effort to further the research on intimate partner homicide. Guided by these three perspectives there are two main research questions that this study hopes to address. Firs t, this research will determine what structural factors correlate with malea nd female-perpetrated intimate partner homicide cross-sectionally (1990 and 2000) before examini ng changes overtime. Next this research will document changes between key structural factors to determine if these changes correlate with the changes in maleand female-perpetrated intimate partner homicide over time. Intimate partner homicide has gone through some remarkable changes over the past couple of decades. Both maleand femaleperpetra ted intimate partner homicide have witnessed drastic declines. However, gender-specific trends in intimate partner hom icide rates show that female-perpetrated intimate partner homicide has declined more drastically than maleperpetrated intimate partner homicide, thus lessening the gap between male and female intimate partner homicide perpetration rate s. Recent research has predomin ately used two main ideas to explain the over two decade dec line in intimate partner homicide the exposure reduction hypothesis and the backlash hypothesis. Research re sults favor both perspec tives. Research also points to the importance of the emergence of domestic violence services and programs with the Battered Womens Movement in the 1970s and the VAWA of 1994. This study is designed to examine these two theoretical perspectives as well as a third theoretical perspective of economic

PAGE 23

23 deprivation and marginalization, which has shown to be very important in the homicide literature (Bailey, 1984; Blau and Bla u, 1982; Land, McCall, and Cohen, 1990; Parker, McCall, and Land, 1999; Steffensmeier and Haynie, 2000a; 2000b; Whal ey and Messner, 2002), in an attempt to gain a better understanding of th e differential gender-specific trends in intimate partner homicide. Structural factors and th e availability of domestic violen ce services are of particular importance. This dissertation begins with Chapter 2 detailing the three theoretical perspectives that are of interest in this stu dy. In Chapter 3 the data and methodology are discussed. Chapter 4 provides the results of the multivariate analysis. Chapter 5 contains conclusions and directions for future research.

PAGE 24

24 0 5 10 15 20 25 1989199019911992199319941995199619971998199920002001YearRate per 100,000 residents Total Homicide Rate Figure 1-1. Total homicide rate per 100,000 residents for large cities, 1989-2001.

PAGE 25

25 0 1 2 3 4 5 6 7 8 9 10 1989199019911992199319941995199619971998199920002001YearRate per 100,000 persons 15 years and older Male-Perpetrated Intimate Partner Homicide Female-Perpetrated Intimate Partner Homicide Figure 1-2. Gender-specific intimate partne r homicide rates for large cities, 1989-2001.

PAGE 26

26 0.00% 5.00% 10.00% 15.00% 20.00% 25.00% 30.00% 35.00% 1989199019911992199319941995199619971998199920002001Year% of Victims Killed by an Intimate Male Female Figure 1-3. Proportion of all homicides involving intimate partners, 1989-2001.

PAGE 27

27 1990 2000 % Change Total Homicide Rate 20.9827 12.1792 41.9 Female-Perpetrated IPH Rate 1.520151 0.730003 51.9 Male-Perpetrated IPH Rate 5.066648 3.325714 34.36 Figure 1-4. Summary of percent change from 1990 to 2000 in total homicide rate, femaleperpetrated intimate partner homicide rate and male-perpetrated intimate partner homicide rate.

PAGE 28

28 CHAPTER 2 LITERATURE REVIEW While research on intim ate partner homicide is growing, the majority of research on intimate partner homicide is directed at describing victimization and/or offending patterns (Block and Christakos, 1995; Browne and Williams, 1993; Gallup-Blac k, 2005; Puzone et al., 2000; Reidel and Best, 1998; Shackelfo rd and Mouzos, 2005) rather th an analyzing these patterns by controlling for the impact of key factors that ma y have an influence on intimate partner homicide over time (Block and Christakos, 1995; Browne and Williams, 1993; Puzone et al., 2000; Reidel and Best, 1998). This research ha s found that the trends in intima te partner homicide during this time period vary not only by the gender of the victim and offender, but also by the type of relationship between the victim and offender. For instance, Puzone et al. (2000) examined U.S. homicide vic timization from 1976-1995 and found that the decline in homicide victimization rates were more pronounced for husbands (75%) than wives (41%), boyfriends (68%) more than girlfriends (21%)6, and ex-husbands (88%) more than ex-wives (41%). Specifically, Puzone et al. (2000) found that the risk of being killed by a spouse in 1995 was 2.6 times greater fo r a wife than a husband, 2 times greater for a girlfriend than a boyfriend, and for the combined period of 1992-1995 was 3.8 times greater for an ex-wife than an ex-husband. Earlier work ex amining maleand female-perpetrated intimate partner homicide supports thes e findings (Browne and Williams, 1993). That is, Browne and Williams (1993) examined Supplemental Homicide Report data from 1976-1987 for all states and found that both married male-perpetrated (i.e., males killing their female spouses) and female-perpetrated (i.e., females killing their male spouses) intimate partner homicide declined, however the decline was more pronounced for the rate of wives killing husba nds than the rate of 6 However, this decline was not statistically significant.

PAGE 29

29 husbands killing wives. In a ddition, Browne and Williams (1993) showed that while marital intimate partner homicide has been decreasing ov er the years, male-perpetrated non-marital intimate partner homicide has actually been increasing. Browne and Williams (1993) concluded that the rate of unmarried females being killed by their intimate partner increased significantly during this period, while the rate that unmarried males were k illed by their intimate partner show no clear trend during the 1976-1987 period (87). Similarly, in Rosenfelds (1997) examination of intimate partner homicide vict imization in California between 1987 and 1996, rates of marital intimate partne r homicide decreased while rates of non-marital intimate partner homicide increased. Rosenfeld ( 1997) attributes this marked ch ange to an increase in the proportion of younger people who are not married. Along these lines, Riedel and Best (1998) also examined intimate partner homicide in Ca lifornia for the years 1987 to 1996 and found that intimate partner homicide is more common in co mmon-law relationships co mpared with spousal and boyfriend/girlfriend relationships They suggest that a reason for this may be because of the meaning that is attributed to the relationship (Makepeace, 1997) and the lack of stability (Makepeace, 1989). In summary, research descri bing recent trends in intimate partner homicide has found some interesting results. It a ppears that while total intimate partner homicide is decreasing, once one disaggregates by gender and relationship type different trends emerge. For instance, overall patterns show that males have experienced a greater decline in intimate partner homicide victimization than females. In addition, married intimate partne r homicide has decreased, while non-married intimate partner homicide has increased and this increase is especially pronounced for non-married female intimate partner victimization.

PAGE 30

30 Though research has not come to a definitive e xplanation for these trends, research that has controlled for the impact of key factors ha s tended to attribute the recent patterns to the social responses toward domestic violence; specificall y, the availability of domestic violence resources (Browne and Williams, 1989; Dugan et al., 1999; Dugan et al., 2003) and have found that the availability of legal (i.e., presence of statutes pertaining to domestic violence) and extralegal services (i.e., number of shelters and other pr ograms) to be related to the decline in the rates of female-perpetrated intimate partner homicide, but not associated with the rates of maleperpetrated intimate partner homicide (Browne a nd Williams, 1989). However, research has also shown that some domestic violence resources (i.e ., prosecutors wiliness to prosecute) that were intended to help women have the unintended co nsequence of making them more at risk for intimate partner homicide victimization (Dugan et al., 2003). Increasing di vorce rates, declining marriage rates, and an improved economic status of women have also shown to be important in explaining intimate partner homicide (Dugan et al., 1999; 2003; Rosenfeld, 1997). These factors will be examined more closely as I describe empi rical findings for the theoretical perspectives of interest in this study. Theoretical Perspectives Three theoretical perspectives are of in terest to this study. The following pages will explain each perspective the exposure reduc tion hypothesis, the backlash or retaliation hypothesis, and economic deprivation and margina lization. The section w ill end with detailed hypotheses for the current study. Exposure Reduction Hypothesis According to the exposu re reduction hypothesi s, limiting the exposure or contact of intimate partners to one another should decrease the probability of intimate partner homicide, because there is less exposure to a violent partne r and more opportunities to exit the relationship.

PAGE 31

31 Ideas behind the exposure reducti on perspective stem from three main findings from previous research. Domestic violence research s hows that intimate partner homicide is commonly the result of prolonged violence in the relationship (Browne, 1987; Campbell, 1992; Goetting, 1995; Smith, Moracco, and Butts, 1998; Totman, 1978) or a prior history of violence (Browne, 1987; Chimbos, 1978; Daniel and Harris, 1982; To tman, 1978). For instance, Campbell (1992) estimated that in Dayton, Ohio 64% of female intimate partner homicide victims were abused physically by their offender before they were ki lled. Also, they estimated that 79% of male intimate partner homicide victims had abused th eir offender previously during the relationship. In addition, self-report studies ha ve shown that abused women make many attempts to try to get outside help before actually committing the intimate partner homicide (Sherman and Berk, 1984). For instance, in reviewing police records in Detr oit and Kansas City, Sherman and Berk (1984) discovered that in 90% of the intimate partner homicide cases police had been called at least 1 time during the 2 years prio r to the homicide. Furthermore, they found that in 54% of the intimate partner homicide cases, police had been called 5 or more times. Other research has shown similar results (Goetting, 1995). Moreover, research has shown that in violent relationships women are more likely to be seriously injured (Berk et al., 1983; Browne, 1993; Brush, 1990; Crowell and Burgess, 1996; Fagan and Browne, 1994; Langan and Innes, 1986; Schwartz, 1987; Stark, Flitcraft, and Frazier, 1979; Stets and Straus, 1990; Stout and Brown, 1995; Straus 1993). In addition, victimprecipitation is more likely in female-perpetrated intimate partner homicide than maleperpetrated intimate partner homicide (Goe tting, 1995; Rosenfeld, 1997; Silverman and Mukherjee, 1987; Wolfgang, 1958). Victim-precipita tion means the victim initiated the incident

PAGE 32

32 that later resulted in their own death by their intimate partner. Rosenfelds (1997) results support this idea as do a number of other research studies (Campbell, 1992; Cazenave and Zahn, 1992; Goetting, 1995; Jurik and Winn, 1990; Silver man and Mukherjee, 1987; Wilson and Daly, 1992a). To summarize, intimate partner homicide is more likely to occur in a relationship where there is a history of violence. In these viol ent relationships, women are more likely to be seriously injured than men. Furthermore, vict im-precipitation is more likely in femaleperpetrated intimate partner homicide. These find ings suggest that wome n have very different motives for killing their intimate partners and wo men are more likely to kill intimate partners out of protection or self-defense (Bernard et al ., 1982; Block and Christ akos, 1995; Browne, 1985; 1987; Campbell, 1992; Daly and Wilson, 1988; Dobash et al., 1992; Dugan et al., 1999; Goetting, 1988; Peterson, 1999; Stout, 1991; Stout and Brown, 1995; see Archer, 2000 and Saunders and Browne, 2000 for a review). That is, wh en women kill intimate partners it is likely a result of attempting to preven t their partners from hurting their children or themselves, during a violent attack by their partner, or in an attempt to prevent a future attack (Browne, 1986; 1987; Dugan et al., 1999). According to Wilson and Daly (1992b) Unlike men, women kill male partners after years of suffering physical violence, after they have exhausted all available sources of assistance, when they feel trapped, and because they fear for their own lives (206). Factors that Reduce Exposure Availability of domest ic violence resources. A main factor that has been shown to influence female-intimate partner homicide by reducing the exposure of violent pa rtners is the availability of domestic violence resources. Browne and Williams (1989) was the first study to examine the influence of the availability of domestic violence resources (i.e., number of shelters; wife abus e programs other than shelters)

PAGE 33

33 and domestic violence legislati on (i.e., statutes providing civil in junction relief for victims of abuse; statutes providing temporary injunction relief during di vorce, separation, or custody proceeding; statutes defining the physical abuse of a family or household member as a criminal offense; statutes permitting warrantless arrest based on probable cause in domestic violence cases; statutes requiring data collection and the reporting of family violence by agencies that serve these families; and statutes providing f unds for family violence shelters) on femaleperpetrated intimate partner homicide. Following prior work on homicide and intimate partner homicide, specifically, (Browne and Flewelling, 1986) Browne and Williams (1989) examined the rate of intimate partner homicide in U.S. states for the year s 1976-1979 and 1980-1984 and found that indeed, domestic violence resources have the ability to provide abused females with alternatives besides killing their male partners Specifically, they found a negative relationship between both domestic violence resources and dome stic violence legislation and intimate partner homicide, however once controls were included in the analysis only the domestic violence resources remained significant. Other research has also supported this hypothesis and shown that social response to domestic violence may be important in explaining intimate partner homicide (Browne and Williams, 1989; Dugan et al., 1999; Dugan et al., 2003). According to Browne and Williams (1989) domestic violence resources provide the means for women (1) to seek protection for them selves and their children, such as emergency restraining orders against abuse and harassment or police removal of the abusive mate; (2) to employ more direct avenues of escape, such as sh elters where they can hide from the abuser for short periods; (3) to ut ilize third-party interventions, such as support groups, crisis counseling, and legal aid, which give advise and encouragem ent in identifying eff ective nonviolent means of responding to threat or danger; and, in some areas, (4) to recei ve the benefits from court-

PAGE 34

34 mandated treatment programs that work directly with the abuser on his problem with violence (76). Dugan et al. (1999) examined the ideas be hind exposure reduction as well, looking specifically at the availa bility of domestic violence resources, as well as declining domesticity (i.e., decrease in the marriage rate), divorce (i.e., increase in the divorce rate) and the improved economic status of women (i.e., fe males educational attainment re lative to males and females income relative to males). They hypothesized that these three factors would reduce the exposure between violent partners. Results are supportive of the exposure re duction perspec tive; however these factors contributed more to the decline in male intimate partner homicide victimization rather than female intimate partner homicide vi ctimization. That is, they found that domestic violence resources (i.e., legal a dvocacy) reduced homicides of ma rried males; however they did not find that domestic violence resources reduced homicides of married females. Dugan et al.s (2003) work was an extension of their previous (1999) work. They again examined the relationship between intimate part ner homicide and domestic violence resources, but they extended their research by including more places (i.e., 48 large cities with a population over 250,000), a longer period of ti me (i.e., 1976-1996), and utilize a richer source of domestic violence resource measures. As with their pr evious research they examined the exposure reduction effects7. They found that aggressive arrest polic ies have an exposure reduction effect such that more aggressive arrest policies in cities is associated with less unmarried intimate partner homicide. In addition, they also showed an exposure reducti on effect such that a decrease in Aid to Families with Dependent Children benef its is related to an increase in male victims killed by their girlfriends, sp ecifically African-American male s. They suggest that public assistance provided to poor women with children may cushion the financial impact of leaving a 7 Dugan et al. (2003) also examined backlash effects of domestic violence resources.

PAGE 35

35 violent partner (Allard et al., 1997), but reductio ns in AFDC benefits would limit opportunities for females to leave their abusive partners a nd live on their own; thus resulting in more opportunities for females to kill their abusive pa rtners. Other variables that were shown to produce an exposure reduction effect include: ma rriage rates, legal advocacy, warrentless arrest laws, and mandatory arrest laws. Declining domesticity. Other common factors that have been cons idered to produce exposure reduction effects are increasing divorce rates, and decreasing marriag e rates. Rising divorce rates would result in fewer married couples living together and woul d therefore reduce the exposure between violent couples. The same idea is behind falling marria ge rates, which would reduce the exposure of violent couples because fewer individuals would be getting married and living together. Rosenfeld (1997) examined intimate partner homici de trends in St. Louis and found that 30% of the decline in African American spousal homicides was attributable to falling marriage rates and rising divorce rates. Dugan et al. (1999) found that domesticity doe s in fact influence intimate partner homicide. They found that the decl ine in marriage rates is related to the decline in married male and female intimate partner victimization. That is they found that the higher the marriage rate in cities, the higher the rates that wives kill th eir husbands and husbands kill their wives. Conversely, a high divorce rate corresponde d to fewer wives killing their husbands8. Specifically, they found that for every reduction of 10,000 women entering marriage, 2.3 married womens lives are saved. The same decrease in marriage among men saves the lives of 3.8 married men andas 10,000 additional men are granted divorce, 3.8 married male homicides are avoided (203). Dugan et al. (2003) found similar results. Specifi cally, they found 8 However, declining marriage rates was related to an in crease in unmarried males killing their female partners.

PAGE 36

36 that lower marriage rates are related to fewer husbands being killed by their wives. Rosenfeld (1997) suggests that much of the decline in intimate partner homicide is a function of change in the rate of marriage with the ag e groups at highest risk of hom icide victimization and offending (73). However, decreasing marriage rates may mean that more individuals are cohabitating together without getting married. Cohabitation has been shown to be an important risk factor in intimate partner homicide (Shackelford and Mouz os, 2005). Research suggests that males and females in cohabitating relations hips are at a higher risk of intimate partner homicide victimization compared to males and females in married relationships (Daly and Wilson, 1988; Shackelford, 2001; Wilson, Daly, and Wright, 1993; Wilson, Johnson, and Daly, 1995). For instance, Wilson, Daly and Wright (1993) and Wilson, Johnson and Daly (1995) found that females that cohabitate with their partner are 9 times more likely to be killed by their intimate partner than are married females. In additi on, Shackelford (2001) found that men that are cohabitating with their female partners are 10 mo re likely to be victims of intimate partner homicide compared to men in married relati onships. Rodriquez and Henderson (1995) suggest that individuals in cohabitating relationships, having invested heav ily in such relationships, may be more likely to resort to violent retaliation (i.e., homicide) than those in dating or noncohabitating relationships because they do not per ceive themselves as having legal protection associated with marriage (48). Improved economic status of women. Improved economic status of women has al so been considered to produce exposure reduction effects. Improved economic status of women in terms of educational attainment, income, and employment increases the access of women to opportunities that may give them

PAGE 37

37 other options besides resorting to killing their male partners. Al so, increased status would reduce females dependence on males. D ugan et al. (1999) did indeed fi nd that females improved status was associated with intimate pa rtner homicide victimi zation, particularly male intimate partner homicide victimization. That is, the increase in females relative income is associated with a decline in married female-perpetrated homicide. Furthermore, an increase in females relative educational attainment is associated with a de cline in non-married ma le victimization. They suggest that more educated women are better able, and perhaps more willing, to exit violent relationships and thus avoid kil ling their partner (204-205). Exposure Reduction Predictions In summ ary, it appears that women in viol ent relationships may be looking for other options, but without other options than remaining in the relationshi p available may turn to lethal methods. Ideas behind the exposure reduction pers pective suggest that limiting the exposure or contact of intimate partners to one another shou ld decrease the probabili ty of intimate partner homicide, because there is less exposure to a violent partner and more opportunities to exit the relationship. Empirically, if exposure reduction efforts were ava ilable to intimate partners before either partner took the violence to a lethal level intimate partner homicide should decrease. Based on these ideas it seems likely that factors that reduce contact or exposure between intimate partners in violent relationships should have a significant impact on female-perpetrated intimate partner homicide, because certain factors may give women other options than having to resort to violence, lethal or not, to protect themselves. The availability of domestic violence resources and the possible effect on intimate partner homicide has gained increase a ttention. Results have been su pportive of the exposure reduction perspective; however this is only true of female-perpetrated intimate partner homicide. It appears that domestic violence services that are designe d to make females safer are actually making

PAGE 38

38 males safer instead. Divorce rates have also proved to be a significant factor influencing intimate partner homicide; as well as cohabitation of intimate partners, and improved status of females. It is believed that measures of exposure reduc tion will have a significant effect on femaleperpetrated intimate partner homicide by creating other options for female s besides killing their male partner. Backlash or Retaliation Hypothesis Although research has shown the importance of reducing the exposure between intim ate partners in violent relationships, it is well known that the highest risk for homicide is when the victim leaves the relationship and this is especially true for females being killed by their male partners (Block and Christakos, 1995; Block, 20 00). Research has suggested the possibility of retaliation by the abusive part ner from domestic violence interventions (Campbell, 1992; Goetting, 1995). Dugan et al. (2003) s findings are supportive of th is statement. Dugan et al. (2003) found a retaliation effect where domestic violence resources actually increased homicide between intimate partners because they failed to effectively reduce e xposure between intimate partners. In fact, the prosecuto rs willingness to prosecute viol ators of protection orders, though intended to reduce exposure between violent intim ate partners, actually caused a retaliation effect where homicide increased for married and unmarried white females and African-American unmarried males. They conclude d that being willing to prosecu te without providing adequate protection may be harmful (192). Contrary to exposure reducti on predictions, research has hypothesized that increased economic status of females would make it easier for females to exit an abusive relationship and may possibly threaten males control over their pa rtners. Dugan et al. (1999) found that net of other changes, more educated women are better able, and perhaps more willing, to exit violent relationships and thus avoid killing their partner (205), howev er improved status appears to

PAGE 39

39 increase the female intimate partner homicide victimization rate. Duga n et al. (1999) suggests that some men are threatened by their loss of power over their female partners as their status increases. The backlash perspective theorizes that incr eased male violence towards their intimate partners is due to a perceived loss of power or control because of womens improving economic conditions (Browne, 1987; Vierai tis and Williams, 2002). This hypothesis was first developed by Williams and Holmes (1981) and Russell (1975) as a warning of the potential consequences of womens increased economic conditions or gender equality during the womens movement. That is, womens improving economic conditions (i.e., increased income, employment, and educational attainment) would make women less dependent on males, which would threaten men and actually cause increased viol ence towards females by males. For instance, Avakame (1999) states initial reductions in gender inequality might cause increases in violence because it would frustrate males into intensifying their use of vi olence to reassert their diminishing patriarchal power and authority (927). Though measures of gender inequality ha ve been identified by feminist scholars as a cause of violence agai nst women, research has shown conflicting findings about the direction of the relationship. Some rese arch has been supportive of this view (Yang and Lester, 1988; Whaley and Messner, 2002), while ot hers have not been (Gartner et al., 1990). Whaley and Messner (2002) hypothesized and found that greater gender equality increased male killings of females, results directly consistent w ith the backlash perspectiv e. They concluded that greater gender equality is threatening to ma le dominance, and as such, it increases male violence against women (199). In addition, other research has shown that when male unemployment is high relative to females, wives killed by their husbands increases (Bailey and Peterson, 1995). However, other research shows contradictory results. Gartner et al. (1990)

PAGE 40

40 examined females and males risk of homici de victimization for the period 1950 to 1980 for 18 developed democracies (i.e., 13 western European nations, Canada, the U.S., New Zealand, Australia, and Japan) and found a relationshi p between higher female education and lower female homicide. They conclude that the fact that the relationship between higher female education and lower female homicide increases over time suggests that the status gained from education acts through incremental changes in po licies and norms proscrib ing violence against women (608). Attempts have been made to test the backlash perspectiv e on other forms of violence, such as rape (Avakame, 1999; Baron and Straus 1987; Ellis and Beatie, 1983), however mixed findings have resulted. For instance, Avakame (1999) examined rape data from the National Crime Victimization Survey (1992-1994) to determin e if females labor force participation had an increase on females rape victimization due to backlash, but did not find support. Instead, Avakame (1999) found that unemployed women we re at a greater likelihood of being raped. However, Ellis and Beattie (1983) showed that gender equality (i.e., in terms of male/female differences in earnings, educational attainment, and employment) is related to an increased likelihood of rape. Surprisingly, Baron and St raus (1987) found the opposite results more gender equality, lower likelihood of rape. Individual-level research has looked at status incompatib ilities between husbands and wives and the influence on physical and emotional abuse. Kaukinen (2004) examined the effect that men and womens relative economic contri butions to the family have on husband-wife physical and emotional abuse. Status compatib ility was broken down into three statuses

PAGE 41

41 traditional status, status parity, and status reversal.9 Kaukinen (2004) found that there was no effect of status compatibility on physical violence, but there was an effect for emotional violence. Specifically, women in status reversal relationships are at risk for emotional abuse by their partner. It seems that men in relationships where their female partners had a higher status exerted their control by emotiona lly abusing their partners. Backlash Predictions Despite the mixed results for the backlash perspective, research has suggested that power differences between partners increase the ri sk for abuse among women (Anderson, 1997). In addition, men who kill their intimate partner do so to dominate and control their female partner (Wilson, 1989; Wilson and Daly, 1992a, 1992b). It is th erefore important to include measures of backlash in the study of maleand female-perpetr ated intimate partner homicide to determine if these factors influence intimate partner homic ide in 1990 and 2000 and whether the changes in these factors contribute to the change s in intimate partner homicide over time. It is expected that measures of backlash will have a significan t impact on male-perpetrated intimate partner homicide, because males will use violence to exert control over their female partners in conditions where female status is high or in co nditions where they feel a perceived lack of control over their female partners. Economic Deprivation and Marginalization The im portance of economic deprivation as a factor influencing crime has been documented in both the strain and feminist literatu res and it is consistent with the other two perspectives of exposure reduction and backla sh. The strain literature posits opportunity structures to be related to crime rate s (Merton [1949], 1968). Merton ([1949], 1968: 185-248) 9 In traditional status the economic status between the woman and the partner favored the partner. In status parity the economic status between the woman and the partner was equa l. In status reversal the economic status between the woman and the partner favored the woman.

PAGE 42

42 argues that deviance is a result of an unequa l distribution of opport unities among classes in society. According to Merton, the distribution of oppor tunities in the social structure determines which classes are most likely to be involved in crime and deviance. Blocked or unavailable opportunities have been suggested to increa se offending (Cohen, 1955; Cloward and Ohlin, 1960; Merton, 1968). Caused from differentials in access to opportunities, economic inequalities in particular have been suggested to increase violence levels (Braithwait e, 1979; Blau and Blau, 1982). Specifically, Blau and Blau (1982) found that relative racial socioeconomic inequalities have a significant influence on criminal violence. That is, aggressive acts of violence seem to result not so much from lack of advantages as from being taken advantage of, not from absolute deprivation but from relative deprivation (126). Other research has examined the influence of economic inequalities on homicide (for review s see Messner and Rosenfeld, 1999). Research has shown that, when economic deprivation is measur ed as an index including many measures of deprivation it is an important predictor of homicide rates (Land, McCall, and Cohen, 1990; Parker, McCall, and Land, 1999). Feminist scholars are interested in a particular form of inequality, particularly genderedinequality and its influence on gendered-violence. Feminist literature signifies the patriarchal system of male domination and control over wome n as a cause of gender inequalities in society and ultimately the oppression of women in society (Walby, 1986). Feminist scholars argue that domestic violence is rooted in gender and pow er and represents mens active attempts to maintain dominance and control over women (Anderson, 1997: 655). When examining gender inequalities a majority of research tests the economic marginaliza tion hypothesis, which attributes female crime to the economic disadva ntage of women relative to men (Heimer, 2000).

PAGE 43

43 Specifically, the economic marginalization hypothesis suggests that as the economic disadvantage of women increases, so will female crime. Research has found support for this hypothesis in regards to female perpetrated o ffenses including intimate partner homicide (Box and Hale, 1983; 1984; Steffensmeier and Streif el, 1992; Dugan et al., 2003; Reckdenwald and Parker, 2008). To test the economic marginaliz ation hypothesis, research has measured the economic disadvantage of females in terms of fe male poverty and joblessness. For example, Box and Hale (1983; 1984) found that female unemployment was positively associated with female violent and property offenses. Impor tantly, research has shown that there is an overlap between factors that effect male and female offe nding (Boritch and Hagan, 1990; Steffensmeier and Haynie, 2000a; Steffensmeier and Haynie, 2000b). For instance, Steffensmeier and Haynie (2000a) examined the effects of structural disa dvantage (i.e., poverty, unemployment, income inequality, female-headed households, and percen t black) on male and female offending rates for multiple offenses (i.e., homicide, robbery, aggravated assault, burglary, larceny-theft). Steffensmeier and Haynie (2000) concluded that structural disa dvantage variables influenced female and male offending rates in similar ways. Surprisingly, the effects of the structural variables were stronger for male offending. The majority of studies have examined gender inequalities in relation to the killings of females (Bailey and Peterson, 1995; Gauthi er and Bankston, 1997; Avakeme, 1999), but examining the impact of gender inequality on male homicides is important as well (Whaley and Messner, 2002). For instance, Whaley and Messner (2002) explored the effects of both gender inequality and economic deprivation on gendere d-homicides (i.e., male offender-male victim, male offender-female victim, female offender-female victim, and female offender-male victim) in large cities for 1990 to 1994. Whaley and Messne r measured gender inequa lity as the ratio of

PAGE 44

44 male to female median income, the percent of males aged 16 and over who were employed relative to the percent of females aged 16 a nd over who were employed, the percent of males employed in the labor force, the percent of managers, executives and administrators who were male, and the ratio of males aged 25 or more w ith 4 or more years of college education to females aged 25 or more with 4 or more year s of college education. They measured economic deprivation as the percentage of black males, pe rcentage of black females, percentage of poor males, percentage of poor females, percentage of unemployed males, percentage of unemployed females, and the gini index of income inequality10. Results suggested that gender equality is positively related to rates of males killing othe r males and males killing females in southern cities. In addition, they found that economic de privation had a significant positive relationship with females killing males, females killing ot her females, males killing females, and males killing other males. Consistent with this research, other re search has found that poverty and economic inequality influence male and female homicide s (Gartner, Baker, and Pampel, 1990; Smith and Brewer, 1992). Also, research has shown that marri ed females are at a higher risk of intimate partner homicide victimization in cities where there are high levels of educational attainment and employment inequalities between men and wo men (Bailey and Peterson, 1995). Furthermore, research has found that womens ec onomic power is significantly re lated to their involvement in intimate partner homicide. Specifically, it has been shown that as womens economic resources increase their involvement in intimate partne r homicide decrease (Gau thier and Bankston, 1997). 10 The gini index represents an index of income inequa lity. The index can range from zero, representing perfect equality, to 1, representing perfect inequality.

PAGE 45

45 Economic Deprivation and Marginalization Predictions Although econom ic deprivation and marginalizat ion is talked about differently in the strain and feminist literatures, it is still appli cable to the current study and should have an impact on gender-specific intimate partner homicide. Economic deprivation is expected to increase both maleand female-perpetrated intimate partner homicide, either in terms of differential opportunity structures or marginalization. Summary In summ ary, three theoretical perspectives are examined to determine their ability in explaining maleand female-intimate partner homici de. Each theoretical perspective is believed to have a unique influence on maleand female-perpetrated intimate partner homicide, based on findings from prior literature. The exposure redu ction perspective and the backlash perspective have shown to be important in understanding in timate partner homicide. Economic deprivation and marginalization has shown to be an importa nt explanation of homicide offending, and is assumed to have an influence on intimate partner homicide as well. Many of the factors believed to represent the theoretical perspectives overlap in many ways. For instan ce, in prior research inequality has been used to represent economic deprivation and/or marg inalization (Whaley and Messner, 2002; Reckdenwald and Parker, 2008), as well as exposure redu ction (Dugan et al., 1999); however in this study inequalit y represents a measure of back lash. I have done my best to break up the variables based on ideas from other re search as well as my research hypotheses for this study. Hypotheses Based on the previous literatu re on intimate partner homicide and homicide in general, it is expected that indicators of exposure re duction, backlash, and ec onomic deprivation and marginalization will significantly impact intimate partner homicide for both males and females

PAGE 46

46 for the years 1990 and 2000. Figure 2-1 contains a de tailed theoretical and conceptual model of the theoretical perspectives on maleand female-perpetrated intimate partner homicide. As illustrated in Figure 2-1, it is believed that m easures of exposure redu ction will significantly decrease female-perpetrated intimate partner homicide, by reducing the exposure between violent partners and reducing the chance that females will have to kill their partners in selfdefense. In addition, it is hypothesized that measures of backla sh will significantly increase male-perpetrated intimate partner homicide, due to males perceived loss of control and dominance over their female partners. Furthermor e, it is believed that measures of economic deprivation and marginalization will increase bot h maleand female-perpetrated intimate partner homicide, whether it is from the l ack of opportunities or marginali zation. Also, it is expected that changes in key indicators from 1990-2000 will imp act changes in maleand female-perpetrated intimate homicide from 1990-2000. Examining these changes will give a cl earer picture of the differential declines witnessed for maleand fema le-perpetrated intimate partner homicide. As depicted in Figure 2-1, I pr opose the following hypotheses: Exposure ReductionH1: Measures of exposure redu ction will ha ve a significant negative impact on femaleperpetrated intimate partner homicide for bot h 1990 and 2000, based on research that suggests that some females are more likely to kill their intimate partner in self-defense and that exposure reduction efforts are mainly geared towards females. BacklashH2: Measur es of backlash will have a significa nt positive impact on male-perpetrated intimate partner homicide for both 1990 and 2000, based on the idea that an increase in the status of females causes some males to kill their female partners in an effort to exert control and dominance over them.

PAGE 47

47 Economic Deprivation and MarginalizationH3: Measures of econom ic deprivation and ma rginalization will have a significant positive impact on both maleand female-perpetrated intimate partner homicide for both 1990 and 2000, due to the linkage in the literat ure of the role that economic de privation and marginalization has on crime. Change ModelsH4: The change from 1990 to 2000 in the theo retical measures of exposure reduction will correlate with the change in female-perpe trated intimate partner homicide from 1990-2000. H5: The change from 1990 to 2000 in the theoretical measures of backlash will correlate with the change in male-perpetrated inti mate partner homicide from 1990-2000. H6: The change from 1990 to 2000 in the theoretical measures of deprivation and marginalization will correlate with the change in both femaleand male-perpetrated intimate partner homicides from 1990-2000. Data is collected for the years 1990 and 2000 to address these hypotheses. The purpose of this study is to address the c ontradictory findings of previous research on intimate partner homicide and reach a better understanding of the tr ends in intimate partner homicide that have been witnessed over this time pe riod. To do this, this study draws on the exposure reduction and backlash perspectives, as well as ideas behi nd economic deprivation and marginalization.

PAGE 48

48 Figure 2-1. Detailed theoretical and conceptual model on exposur e reduction, backlash, economic deprivation and marginalization and maleand female-perpetrated intimate partner homicide. Female-Perpetrated Intimate Partner Homicide Male-Perpetrated Intimate Partner Homicide + + + Exposure Reduction % Divorced % Unmarried Households Shelter Rate Legal Assistance Rate Batterers Counseling Rate Referral Rate Economic Deprivation % Living in Poverty % Unemployed % HHs on Public Assistance Backlash (higher levels of equality) Ratio F/M Median Income Ratio F/M Education Ratio F/M Employment

PAGE 49

49 CHAPTER 3 DATA AND METHODOLOGY This se ction details the methodological strategy used in this study. To begin, the data sources and the unit of analysis are discussed. Next, all measures are operationalized. After examining the data three methodological issues are apparent. First, there is a number of missing cases in the dataset in regards to offender info rmation and domestic violence resource information. Techniques for dealing with this will be presented. Second, ther e is evidence of collinearity between predictors included in this study. To address this concern, principal components (specifically principal axis factoring) analys is is employed. Third, Ordinary Least Squares regression estimations are inappropriate due to th e skewed nature of the dependent variables. Poisson-based regression techniques will be utilized for the cross-sectional models, because the dependent variables are based on di screte counts of rare events and have a skewed distribution. These issues will be outlined in the followi ng pages. Unit of Analysis For this study, the unit of an alysis is cities with a popul ation of 100,000 or m ore in 1990. This definition of urban cities is consistent with the Bureau of Justice Statistics definition of large cities (Bureau of Justice Statistics 2006). The resulting urba n sample contains 200 cities that meet that criterion. Of those cities 178 cities were used in the analysis. Chattanooga, TN, Citrus Heights, CA, East Los Angeles, CA, Kansas City, KS, Lowell, MA, Metairie, LA, Omaha, NB, Overland Park, KS, Paradise, NV, Springfield, IL, St erling Heights, MI, Worchester, NY, and all cities in the state of Florida were droppe d from the analysis due to not reporting at all during either the period 1989-1991 or 1999-2001. Chicago, IL was the final city dropped due to being an extreme outlier. The average city population in 1990 is 323,445 ( =

PAGE 50

50 629,308), ranging from 100,217 to 7,322,564. The averag e city population in 2000 is 355,578 ( = 686,124), ranging from 95,658 to 8,008,278. Sources of Data Multip le data sources were used for the current study. The dependent variables are based on data from the Supplemental Homicide Files for the years 1989-1991 and 1999-2001 (Fox, 2005). U.S. Bureau of Census summary files 111 and 312 for 1990 and 2000 are the sources of data for the independent variables. Uniform Crime Reports: Police Employee (LEOKA) Data are the source of data for police force size (cons istent with work by Steffensmeier and Haynie 2000a; 2000b). Also, the 1991 and 1999-2000 Domestic Violence Service Directory, collected by the National Coalition Against Domestic Vi olence, is the source for the data on the availability of domestic violence services. These years were chosen because they were the years published by the National Coalition Against Do mestic Violence and represented 1990 and 2000 closest. Homicide Data The hom icide data used in this study were obtained from the Supplemental Homicide Reports (SHR). The SHRs are part of the Unif orm Crime Reporting Program (UCR). In addition to monthly criminal offense information compile d for UCR purposes, law enforcement agencies submit supplemental data to the Federal Bureau of Investigation on homicide incidents. The SHRs contain detailed, incident-level data on nearly all murders and nonnegligent manslaughters that have occurred in the United States for a given year. SHRs contain information for each 11 Summary File 1 presents 100percent population and housing data for the total population, for 63 race categories, and for many other race and Hispanic or Latino categories. See Appendix A for a more detailed definition. 12 Summary File 3 presents data on population and housing long-form subjects, such as income and education. See Appendix A for a more detailed definition.

PAGE 51

51 homicide incident, including info rmation on victim and offenders gender as well as victim and offender relationship information. Domestic Violence Service Data The data on dom estic violence services were obtained from the 1991 and 1999-2000 Domestic Violence Service Directory, collected by the National Coalition Against Domestic Violence. Unfortunately, the National Coalition Ag ainst Domestic Violence does not collect data ever year. These years were chosen because they represented 1990 and 2000 the closest. The information about each domestic violence program for each city is provided by the domestic violence programs themselves. For 1999-2000, th e National Directory includes information on whether 16 service categories are provided by the domestic violence program in each city. These service categories include: sh elter, safe homes, 24-hour hotlines, TTY/TDD (ability to communicate by telephone with women w ho are hearing impaired or deaf), counseling/advocacy, non-resident support groups childrens counseling/programs, rape/sexual assault services, emergency tran sportation, legal services/advocac y, wheelchair accessible, age limit for male children, counseling for batterers, shelter capacity, maximum length of stay, and languages other than English spoken. For 1991, th e National Directory includes information on whether 14 service categories are provided by the domestic violence program in each city. These service categories include: sh elter, safe homes, 24-hour hot lines, accept collect calls, counseling/advocacy, non-resident support groups childrens counseling/programs, fees for service, transportation provide d, legal services/advocacy, coun seling for batterers, wheelchair accessible, deaf services, and languages spoken besides English.

PAGE 52

52 Measures Dependent Variable The Supplem ental Homicide Reports (SHR) ar e the data source for the two dependent variables maleand female-perpe trated intimate partner homicide counts of arrest for the years 1990 and 2000. To determine the counts of maleand female-perpetrated intimate partner homicide for a given city and a given year, the counts were aggregated to the city-level for each year by the offenders gender and intimate partne r relationship. Consistent with other work (Messner et al., 2005; Villarreal, 200 4) arrest counts for each city are based on three year sums (1989, 1990, 1991; and 1999, 2000, 2001) to control for any fluctuations in reporting and relative low frequency of offending. For the current study only murd ers and nonnegligent manslaughters with single victim and single offender were incl uded. This is consistent with other research (Flewelling and Williams, 1999; Williams and Flewelling, 1987). Also, intimate partners include spouses, common-law spouses, ex-spouses, and boyfriends/girlfriends. Homosexual relationships were excluded for the purpose of this study, b ecause the male-female dynamic is of particular importance and because of the small number of cases (Dugan et al., 2003). Independent Variables U.S. Bureau of Census data for the year s 1990 and 2000 were used as information for the majority of the independent variables. Police Em ployee (LEOKA) Data were the source of data for the officer rate. For simplicity, in the followi ng section the independent variables are broken down into 4 categories of measures: exposure reduction, backlash, economic deprivation and marginalization, and control variables. Gender-speci fic indicators are utilized when appropriate. It is acknowledged that many of these measures may overlap and could possibly fall within another category of measures. Fo r the purpose of this study the m easures are divided with the

PAGE 53

53 purpose of capturing the three theoretical perspe ctives of exposure reduction, backlash, and economic deprivation an d marginalization. Exposure reduction measures To assess whether domesticity and domestic vi olence service availability had an exposure reducing effect on intimate partne r homicide, eight measures of exposure reduction are included in this study. These measures in clude: percentage of divorced males, percentage of divorced females, percentage of unmarried partner house holds, the number of shelters offered per 100,000 females, the number of legal services offe red per 100,000 females, the number of referrals offered per 100,000 females, and the number of male batterers counseling services offered per 100,000 males. Percentage of divorced males. The first meas ure of exposure reduction is the percent of males that are divorced. This was calculated by dividing the number of ma les aged 15 and older who are divorced by the total number of males ag ed 15 and older. This calculation was then multiplied by 100 to obtain a percent. 100 age) of years 15 Males No.( divorced) are that age of years 15 Males No.( Percentage of divorced females. The percent of females that are divorced was calculated the same as the percent of males divorced. The number of females aged 15 and older who are divorced was divided by the total number of females aged 15 and older and then multiplied by 100 to obtain a percent. 100 age) of years 15 Females No.( divorced) are that age of years 15 Females No.( Percentage of unmarried partner househol ds. Another measure of exposure reduction is the percent of unmarried partner households. Th is measure captures cohabitation by unmarried

PAGE 54

54 partners. Unmarried partners incl ude male householders living with a female partner and female householders living with a male partner. This m easure was calculated by dividing the number of unmarried partner households by the total numbe r of households and then multiplying this by 100 to obtain a percent. 100 ) Households no. Total( ) Households Partner Unmarried No.( Shelter rate. To measure the rate of the numbe r of shelters available to females a measure of the number of shelters for each city and year per 100,000 females 15 years and older was computed. The number of shelters was divided by the total number of females aged 15 years and older and then multiplied by 100,000 to obtain a rate. This is similar to Dugan et al. (1999)s calculation for the availability of hotli nes, counseling, and legal services. older) and years 15 aged females ofNumber ( year) andcity each for shelters of(Number x 100,000 Legal service rate. The legal service rate is de signed to measure the availability of legal assistance, such as in obtaining restraining orders, court accompaniment, legal clinics or advocacy. To measure the rate for the number of le gal services available to females a measure of the number of legal services fo r each city and year per 100,000 females 15 years and older was computed. The number of legal services was di vided by the total number of females aged 15 years and older and then multiplied by 100,000 to obtain a rate. older) and years 15 aged females ofNumber ( year) andcity each for services legal of(Number x 100,000 Number of referrals rate. Many cities may not have offere d a particular service (i.e., shelters, hotlines, counseling, childrens counseling, and legal services) but were able to refer individuals to other programs that offered the service. To capture this, a measure for the rate of

PAGE 55

55 the number of referrals per 100,000 females aged 15 and older for a particular city offered was calculated. The number of referrals was divided by the total numb er of females aged 15 years and older and then multiplied by 100,000 to obtain a rate. older) and years 15 aged females ofNumber ( year) andcity each for referrals of(Number x 100,000 Male batterers counseling rate. To measure the rate for the number of male batterers counseling services available to males a measur e of the number of male batterers counseling services for each city and year per 100,000 ma les 15 years and older was computed. The number of batterers counseling services was divided by the total numbe r of males aged 15 years and older and then multiplied by 100,000 to obtain a rate. older) and years 15 aged males ofNumber ( year) andcity each for services counseling batterers ofumber ( N x 100,000 Backlash measures Three ratio measures of female status rela tive to men were incl uded in this study to capture the ideas behind backlash. These measures have been used in previous research (Whaley and Messner, 2002). They include: the ratio of male to female median income, the ratio of the percentage of male employment to the percentage of female em ployment, and the ratio of the percentage of male education to the percentage of female education. A value of 1.0 on the ratio measures would represent equality between males and females. Values greater than 1.0 represent greater disadvantage of females to males. Ratio of male to female median income. The first backlash measure is the ratio of male to female median income. This measure was calculated by dividing the male median income by the female median income.

PAGE 56

56 income) median Female( income median Male( Ratio of the percentage of male employment to the percentage of female employment. The second backlash measure is the ratio of the percentage of male employment to the percentage of female employment. This measur e was calculated by dividing the percent of males aged 16 or older who were employed in the civilian labor force by the percent of females aged 16 or older who were employed in the civilian labor force. force)labor civilian in employed are whomoreor 16 aged females of %( force)labor civilain in the employed are whomoreor 16 aged males of %( Ratio of the percentage of males with college education to the percentage of females with college education. The final backlash measure is the ratio of the percentage of males with a college education to the percentage of female s with a college education. This measure was calculated by dividing the percent of males aged 25 and older with 4 or more years of college education by the percent of females aged 25 and ol der with 4 or more years of college education. education) college of years moreor 4 older with and 25 aged females of %( education) college of years moreor 4 older with and 25 aged males of %( Economic deprivation and marginalization measures Two gender-specific measures were incl uded in the models to capture economic deprivation and marginalization. These measur es included: percentage of males living in poverty, percentage of females living in poverty, percentage of males unemployed, and the percentage of females unemployed. A third meas ure of the percentage of households on public assistance was also included to measure economic deprivation. Percentage of males living in poverty. The first measure is the percentage of males living in poverty. It was calculated by dividing the nu mber of males below the officially defined

PAGE 57

57 poverty threshold in either 1989 or 1999 by the total number of males for whom poverty status was determined in 1989 or 1999 and then multiplied by 100 to obtain a percent. 100 1989/1999) in ed deterermin wasstatus poverty for whom Females Total( )level poverty below incomes with 1989/1999 in determined wasstatus poverty for whom Females (No. Percentage of females living in poverty. The next measure is the percent of females living in poverty. It was calculated by dividing the numb er of females below th e officially defined poverty threshold in 1989 or 1999 by the total number of female s for whom poverty status was determined in 1989 or 1999 and then multiplied by 100 to obtain a percent. 100 1989/1999) in ed deterermin wasstatus poverty for whom Males Total( )level poverty below incomes with 1989/1999 in determined wasstatus poverty for whom Males (No. Percentage of males unemployed. Another measure represen ting economic deprivation and marginalization is the percent of male s unemployed. This measure was calculated by dividing the number of males 16 yrs and older who are unemployed by the total number of males aged 16 and older and then multiplying this by 100 to obtain a percent. 100 age) of years 16 Males No.( ) unemployed are that age of years 16 Males No.( Females unemployed. Another measure representi ng economic deprivation and marginalization is the percent of females unemp loyed. This measure wa s calculated by dividing the number of females 16 years and older who are unemployed by the total number of females aged 16 and older and then multiplying this by 100 to obtain a percent. 100 age) of years 16 Females No.( ) unemployed are that age of years 16 Females No.( Percentage of households on public assistance. The final measure representing economic deprivation and marginalization is the percent of households on public assistance income. This

PAGE 58

58 measure was calculated by dividing the number of households on public assistance income by the total number of households and then multiplying this by 100 to obtain a percent. 100 ) Households of No.( ) assistance publicon Households of No.( Controls Four control measures were included in this study. They include: residential mobility, officer rate, log of the percen t Hispanic, and the south. Residential mobility. Residentia l mobility (i.e., percent of the population who have lived in current household for 5 years or less) was included in this study to represent a standard social disorganization measure. It was measured by dividing the number of pe ople in the population 5 years and older that had lived in a different house 5 years prior by the total population 5 years and older and then multiplying this by 100 to obtai n a percent. This variable has been suggested to influence intimate partner homicide (Brown e and Williams, 1989; Williams and Flewelling, 1988), particularly female-perpetrated intimate pa rtner homicide as a result of male aggression. Browne and Williams (1989) found that disruptions in life in the form of population mobility were associated with greater female-perpetrated intimate partner homicide. They suggest that mobility may isolate females from support networks that would gi ve females other avenues when conflicts arise instead of resorting to lethal violence. 100 age) of years 5 Population( 1995/1985) in house different ain lived that age of years 5 Population ( Officer rate. The officer rate was calculate d by dividing the total number of officers for each city by the total population and then this wa s multiplied by 1,000 to obt ain a rate of officers per 1,000 persons in the population. This calculation is consistent with other work (DeWees and Parker, 2003; Steffensmeier and Haynie, 2000a).

PAGE 59

59 Log Percent Hispanic. Another control variable is the percent Hispanic population. This variable is included in the models to account for the fact that research has shown that Hispanics are less likely to be involved in intimate partner homicide (Pau lsen and Brewer, 2000). In their analysis of spousal sex ratios of killing (i.e., the number of fe male intimate partner homicide perpetrators for every 100 male intimate partner homicide perp etrators), Paulsen and Brewer (2000) found that the largest disp arities between male-perpetrated intimate partner homicide and female-perpetrated intimate partner homicide was for Hispanics. One reason could be that domestic violence and homicide is viewed differently in the Hispanic culture. To calculate the percent Hispanic the total Hispanic population wa s divided by the total population and multiplied by 100 to obtain a percent. The natu ral logarithm of the percent Hisp anic was then calculated due to the skewed nature of the independent variable. 100 )Population Total( )population Hispanic Total(Ln South. A dummy coded variable for th e south is used in the present analysis to control for any regional differences in intimate partne r homicide offending. Research has shown that homicide rates are higher in the south (Nisbett and Cohen, 1996) and th at partner homicide is higher in the south as well (Browne and W illiams, 1989; Nisbett and Cohen, 1996). This measure is based on Census Bureau definitions of southern regions (i.e., coded as 1 for cities located in the southern region and 0 for ci ties not located in the southern region). Missing Data Missing information in the homicide data was problematic. For the most part this is caused by the fact that partic ipation in the Supplemental Ho micide Reports program is 1,000 1990/2000) in population Total( officers) police ofnumber (Total

PAGE 60

60 completely voluntary. Therefore, some law enfor cement agencies fail to report their homicide incidents to the FBI. According to the Bureau of Justice Statistic s (2006) the SHRs are just over 90% complete. Though the coverage is high, ther e are still a number of homicides that go unaccounted for. This underreporting by law enfo rcement agencies was corrected with an adjustment factor based on the to tal number of homicide incidents reported to the Uniform Crime Reporting program (see Fox, 2004 for discussion on weight calculation; Williams and Flewelling, 1987). That is, SHR records were adjusted so that State and national total counts of murder and nonnegligent manslaughter matched UCR es timates. It is importan t to note that this weighting process assumes that the missing record s are not systematically different from those that have been reported to the FBI. In additi on, there is a problem with using offender data because there is a growing problem with unsol ved murders. Ignoring homicides with missing offender information understates homicide offend ing. A weighting strategy based on available information about the victims murdered in both solved and unsolved homicides is used in the present study to adjust for missing offender data (see Fox, 2004 for disc ussion). In the current analysis all cases are weighted to reflect U CR estimated U.S. homici de counts and imputed offender data. Missing data were also an issue with the domestic violence service data. There were instances where a program in a city did not report whether or not a particul ar service was offered. In these cases, the mean of that particular se rvice across cities was imputed. For the domestic violence service variables included in the mode l, a dummy variable for whether the mean was imputed for a city was included in the final m odels (i.e., 1 = mean imputation, 0 = no mean imputation for a city) to determine if the missing values were missing at random. Results showed

PAGE 61

61 that the missing values were missing at random and therefore the mean imputations were appropriate to use. Multicollinearity There are two ways to assess mulitcollinearity: (1) examining bivariate correlations between the variables and, (2) examining varian ce inflation factors. The bivariate correlation matrices for all variables in 1990 and 2000 are pr esented in Appendix B and Appendix C. After examination of the correlation matric es it is evident that there is collinearity and partiality among the regressors. While there is no clear standard, researchers generally ac cept a correlation greater than .500 to mean that there may be problems with mulitcollinearity. Examination shows that there are correlations between the independent va riables that are above .500. For instance, in 1990 and 2000 the percent of female unemployment and the percent of females living in poverty are highly correlated (r =.782 and r = .816 respectiv ely). The percent of females living in poverty and the percent of households on pu blic assistance are co rrelated (r = .786 and r = .751). The percent of females unemployed and the percent of households on public assistance income are also correlated (r = .793 and r = .751). Furtherm ore, the percent of male unemployment and the percent of males living in povert y are highly correlated (r=.786 and r = .788 respectively). The percent of males living in pove rty and the percent of househol ds on public assistance are correlated (r = .766 and r = .757). The percen t of males unemployed and the percent of households on public assistance income are also correlate d (r = .863 and r = .816). To be sure that multicollinearity is not a problem variance inflation factors were calculated. Generally it is accepted that VIFs gr eater than 4 represent multicollinearity (Fisher and Mason, 1981:109). However, some have sugges ted that a value of 10 or greater indicates multicollinearity (Ott and Longnecker, 2001:652 ; Pindyck and Rubinfeld, 1998). Variance inflation factors for all indepe ndent variables are presented in Table 3-3. For the female-

PAGE 62

62 perpetrated intimate partner homicide model in 1990, the percent of females living in poverty, the percent of females unemployed, and the perc ent of households on public assistance have a variance inflation factor greater than the value of 4 (5.465, 4.887, 5.192, respectively). None of the other independent variables have a variance inflation factor grea ter than 4. As in the femaleperpetrated intimate partner homicide model, the same three variables have variance inflation factors greater than four th e percent of males living in poverty, th e percent of males unemployed, and the percent of households on public assistance (4.697, 6.970, 4.871, respectively). Again, none of the other indepe ndent variables have a variance inflation factor greater than 4. For the female-perpetrated intimate partner homicide model in 2000, the percent of females living in poverty and the percent of households on public assistance have variance inflation factors that are greater than the valu e of 4 (4.827 and 4.496 respectively). In addition, the percent of females that are unemployed has a vari ance inflation factor that is very close to the value of 4 (3.834). None of the ot her independent variables have a variance inflation factor close to or greater than 4. For the male-perpetrated intimate partner homicide model in 2000, the percent of males living in pove rty, the percent of males une mployed, and the percent of households on public assistance all have variance inflation factors greater than 4 (4.040, 4.494, and 4.865 respectively). As with the female-perpetr ated intimate partner m odel, none of the other independent variables have a variance inflat ion factor close to or greater than 4. Due to relatively high variance inflation fact ors, the percent living in poverty, the percent unemployed, and the percent of households li ving on public assistance were included in a confirmatory factor analysis to determine if these variables loaded together. An index was developed using principle axis factoring with varimax rotation. The principal components

PAGE 63

63 analysis for 1990 is displayed in Table 3-4 and th e analysis for 2000 is presented in Table 3-5. The measures of percent unemployment, percent poverty, and the percent of household living in poverty loaded together in both the male-perpetrated and female -perpetrated models for both time periods. An index was create d that is referred to as the male economic deprivation index and the female economic deprivation index. The simplified correlation matrices for 1990 and 2000 are presented in Appendix C and Appendix D. In addition, variance inflation factors for the simplified models are disp layed in the Appendix F.13 The bivariate correlation matrices for the domestic violence service variables clearly show that many of the individual domestic violen ce services are highly collinear with each other (See Appendix G). Four variables were of particular interest because th ey represented service information that were collected for both time periods and contained the most information. These variables included whether or not shelters, legal services, male ba tterers counseling, and referral services were offered for a city.14 In the final analyses composite measures of these variables are utilized. Examining variance inflation factors revealed that collinearity was not an issue between these variables. Methodology Descriptive Statistics The means and standard deviations for all th e variables in 1990 and 2000 can be seen in Table 3-1. Examining the dependent variables shows that both maleand female-perpetrated 13 The variance inflation factors for the simplified male-per petrated and female-perpetrated models in 1990 and 2000 reveal that after principal components is considered and the index made, the variance inflation factors do not exceed 4.0. 14 Initially an index was created that measured the availability of domestic violence services, which included the number of programs per 100,000 females offered for each city and the number of different domestic violence services offe red (i.e., whether the domestic violence program provided shelters, a 24-hour hotline, counseling/advocacy, childrens counseling, and legal/services/advocacy) pe r 100,000 females. The index for the availability of domestic violence services did not obtain si gnificance in the models. However, once the services were measured separately individual services reached significance. Therefore, indivi dual services were included in the final analysis, instead of the combined index.

PAGE 64

64 intimate partner homicides decreased from 1990 to 2000. The mean of female-perpetrated intimate partner homicides (i.e., 3 year sum of counts) in 1990 is 6.553 ( = 10.864). In comparison, the mean of female-perpetrated intimate partner homici des in 2000 is 2.970 ( = 4.779). The mean of male-perpetrated intim ate partner homicides in 1990 is 11.569 ( = 18.378) and the mean in 2000 is 8.770 ( = 13.929). For the indicators of exposure reduction, both the percent of males divorced and females divorced decreased from 1990 to 2000, whereas th e percent of unmarried partner households increased. In addition, the rate of the number of shelters available to females and the rate of the number of legal services avai lable to females increased from 1990 to 2000 (.354 to 1.49 and .313 to 1.42, respectively). In contrast the rate of the number of male batterers counseling programs and the rate of the number of referral services offered decreased from 1990 to 2000 (.746 to .552 and .089 to .075, respectively). Examining the indicators of backlash shows that the disparities between male and female educational attainment and empl oyment decreased (1.33 to 1.14 and 1.24 to 1.19, respectively), whereas the disparity be tween male and female income increased from 1990 to 2000 (1.42 to 1.55). This suggests that over time, the gap betw een males and females educational attainment and employment are becoming more narrower, while at the same time the gap between males and females median income is widening. For the indicators of economic deprivation a nd marginalization, both the male and female economic deprivation indices decreased fr om 1990 to 2000. In 1990 the mean of the female economic deprivation index was 29.38 ( = 11.97) and in 2000 it was 15.88 ( = 6.22). In 1990 the mean of the male economic deprivation index was 26.97 ( = 11.28) and in 2000 was 14.88 ( = 5.88).

PAGE 65

65 Table 3-2 reports the paired sample t-test values to show whether the variable means are significantly different between 1990 and 2000. All the variable m eans are significantly different except for the mean of the num ber of referral serv ices per 100,000 females. In summary, statistics show that both maleand female-perpe trated intimate partner homicide decreased from 1990 to 2000. In addition, for both males and fe males, economic depriv ation and the percent divorced decreased from 1990 to 2000, as well as the ratio of male to females educational attainment, the ratio of male to female employm ent, and the rate of male batterers counseling programs. In contrast, the ratio of male to female median income, percent of unmarried households, the rate of the number of shelters, the rate of the nu mber of legal services, percent Hispanic, residential mobility, and the officer rate increased from 1990 to 2000. Analytical Plan Cross-Sectional Analyses Cross-sectional regression m odels are used to analyze the impact of the theoretical concepts at two points in time (1990 and 2000). Po isson regression models will be used for the current study, because the dependent variables ar e based on discrete counts of rare events and have a skewed distribution. In addition, the depe ndent variables include a number of cities with zero counts, all of which make OLS inappropriate to use (Osgood, 2000; Osgood and Chambers, 2000). The poison regression model is the simplest re gression model for count data. However, in the poison regression model the variance is forced to equal the mean, which is often not the case with count data. After examining the mean and vari ance of the current dist ribution it is clear that the mean and the variance are not equal (Cameron and Trivedi, 1998). The negative binomial regression model is more appropriat e to use because it allows for overdispersion of the data (i.e.,

PAGE 66

66 where the variance exceeds the mean) (Gardne r et al., 1995; Osgood, 2000). In the negative binomial distribution the mean and the variance are as follows: )/( )Var( )(2k [1/k is the dispersion parameter] Though negative binomial models take into a ccount overdispersion, they usually under predict the number of zero counts in the dependent variable. Since both maleand femaleperpetrated intimate partner homicides are rare ev ents and therefore result in a number of zero counts across cities, zero-inflated negative binomia l regression is more appropriate and will be utilized. Furthermore, the vuong test produced a z-value of 14.55 (p = .000) for the maleperpetrated intimate partner homicide model and a z-value of 15.05 (p = .000) for the femaleperpetrated intimate partner homicide model. A significant value means that the zero-inflated model is a better fit of the data. Excess zero c ounts may cause overdispe rsion and zero-inflated models take this into consideration (Cameron and Trivedi, 1 998). In both models the intimate partner homicide count was converted into an equivalent of a rate by including the logged gender-specific population size multiplied by a factor of 3 (i.e., to account for the 3 year pooled dependent variable counts) as an exposu re variable in the model (Agresti, 1996). Pooled Cross-Sectional Time Series Analysis: Fixed Effect Estimation Using STATA, version 9.1, pooled cross-sectio nal time series analyses with fixed effect estimation are used to estimate the impact of ke y theoretical constructs on changes from 1990 to 2000 in maleand female-perpetrated intimate pa rtner homicide. An Ordinary Least Squares fixed effects regression15 was employed as follows 15 Negative binomial fixed effects and random effects regression models were estimated, due to the large number of zero counts and skewed distribution fo r the dependent variables. However, the model for the female-perpetrated intimate partner homicide data would not allow for the es timation of the Hausman test because the data failed to

PAGE 67

67yit = xit + i + t + it i = 1, N; t =1, T where xit is the vector of covariates for both cities (i) and time periods (t), with dummy variables controlling for city (i) and time trends ( t). A time variable for year 1990 was included in the analysis to control for period effects, with the year 2000 as the reference period. There are advantages to this type of analysis. For instan ce, pooled cross-sectional time series analysis increases the number of observations (i t) and also allows one to model time and space and generalize between them. Due to the skewed nature of the dependent variables both the female-perpetrated intimate partner homicide count over 3 years and the male-perpetrated intimate partner homicide count over 3 years was transformed using a logarithmi c transformation. An issue arose with the logarithmic transformation when the count of inti mate partner homicide equaled zero. The log of a zero equals To fix this problem, all zero counts we re also coded as zero for the logarithmic transformation of the intimate partner homicide count. To account for this a dummy variable was included in the models that represented zero coun ts for intimate partner homicide (See Hausman, Hall, and Griliches, 1984 for rationale). Also, the log of the population 15 and older multiplied by a factor of 3 was included in the m odels to control for the population size. Both fixed-effects and random-effects models are appropriate to use for pooled crosssectional analysis, however for the current study the fixed effect estimation was used for a meet the asymptotic assumptions of th e Hausman test. The assumptions of th e Hausman specification test are often hard to meet. The Hausman test assumes that one of th e estimators is efficient (i.e., has minimal asymptotic variance). This assumption is violated if observations are clustered or pweighted or if the model is misspecified (STATA 2008). For comparison purposes, both female-perpetrated and male perpetrated intimate partner homicide were modeled using OLS fixed effects regression.

PAGE 68

68 variety of reasons: (1) it expre sses variables only in te rms of change within a unit (Parker, 2004), which is appropriate because the current study seeks to determine the relationship between change in key theoretical construc ts and change in the dependent va riable for a given city; (2) it controls for all omitted and included time invari ant variables (Allison, 1994; Johnson, 1995); and (3) as described below, the Hausman specification test indicated that the fixed effects model was more appropriate to use than the random effects model. The results of the Hausman specification test (1978) revealed that fixed effects estimation was more appropriate than random effects estima tion for both the male-perpetrated and femaleperpetrated intimate partner homicide multivariate models. The Hausman specification test examines the null hypothesis of no difference be tween the parameter estimates of the more efficient random effects model and the fixed effects model, whereas the alternative hypothesis states that there is a significant difference be tween the random effects models parameters and the fixed effects models parameters. This difference indicates bias in the random effects estimation and thus the fixed e ffects estimation is preferred (Greene, 2000; Hausman et al., 1984; Hsaio, 1986; Maddala, 1983). The Hausman specification test pr oduces a Chi-square statistic, such that a significant p-value indicates that fixe d effects estimation is more appropriate and an insignificant p-value indicates that random effects estimation is more appropriate.

PAGE 69

69 Table 3-1. Means, standard deviat ions (in parentheses) for all vari ables, 1990 and 2000 (N= 178). 1990 2000 Female-perpetrated intimate partner homicide 6.553(10.864) 2.970(4.779) (counts summed over 3 years) Male-perpetrated intimate partner homicide 11.569(18.378) 8.770(13.929) (counts summed over 3 years) Percent female divorced 13.31(4.24) 12.25(2.08) Percent male divorced 9.92(3.12) 9.31(2.03) Percent unmarried households 3.91(1.54) 5.25(1.068) Male batterers counseling per 100,000 males .746(1.09) .552(.738) Shelters per 100,000 females .354(.276) 1.49(.909) Legal service programs per 100,000 females .313(.289) 1.42(1.167) Referral services per 100,000 females .089(.243) .075(.187) Ratio of male to female education 1.33(.140) 1.14(.098) Ratio of male to female income 1.42(.173) 1.55(.180) Ratio of male to female employed 1.24(.106) 1.19(.086) Female economic deprivation Index 29.38(11.97) 15.88(6.22) Male economic deprivation index 26.97(11.28) 14.88(5.88) Percent hispanic 13.40(15.67) 18.61(18.17) Residential mobility 50.22(7.88) 51.40(6.71) Officer rate per 1,000 1.98(.910) 2.22(1.04) South .303(.461) .303(.461)

PAGE 70

70 Table 3-2. T-test scores for change from 1990 to 2000 (N=178). Mean 1990 Mean 2000 Mean Diff t Female-perpetrated intimate partner homicide 6.55(10.86) 2.97(4.78) -3.58 -5.83** Male-perpetrated intimate partner homicide 11.57(18.38) 8.77(13.93) -2.80 -4. 32** Percent female divorced 13.31(4.24) 12.25(2.08) -1.06 -3.45** Percent male divorced 9.92(3.12) 9.31(2.03) -.61 -2.76** Percent unmarried households 3.91(1.54) 5.25(1.07) 1.34 12.17** Male batterers counseling per 100,000 males .75(1.09) .55(.74) -.20 -2.99** Shelters per 100,000 females 35(.28) 1.49(.91) 1.14 18.11** Legal service programs per 100,000 females .31(.29) 1.42(1.17) 1.11 13.29** Referral services per 100,000 females .09(.24) .08 (.19) -.01 -.695 Ratio of male to female education 1.33(.14) 1.14(.10) -.19 -28.94** Ratio of male to female income 1.42(.17) 1.55(.18) .13 12.09** Ratio of male to female employed 1.24(.11) 1.19(.09) -.05 -5.97** Female economic deprivation Index 29.38(11.97) 15.88(6.22) -13.5 -13.64** Male economic deprivation index 26.97(11.28) 14.88(5.88) -12.09 -2 4.85** Percent hispanic 13.40(15.67) 18.61(18.17) 5.21 14.03** Residential mobility 50.22(7.88) 51.40(6.71) 1.18 2.76** Officer rate 1.98(.910) 2.22(1.04) .24 4.74** + p < .10 p < .05 ** p < .01

PAGE 71

71 Table 3-3. Variance inflation factors for all variables included in the models (N=178). F-perp90 M-Perp90 F-perp00 M-perp00 Percent female divorced 1.228 1.464 Percent male divorced 1.204 1.666 Percent unmarried households 1.366 1.360 2.830 2.940 Male batterers programs per 100,000 males 1.245 1.241 1.405 1.399 Shelters per 100,000 females 2.198 2.138 2.392 2.387 Legal service programs per 100,000 females 2.101 2.129 2.874 2.861 Referral services per 100,000 females 1.179 1.163 1.300 1.323 Ratio of male to female median income 1.883 1.818 2.110 2.159 Ratio of male to female education 1.580 1.550 2.080 2.120 Ratio of male to female employment 1.443 1.468 2.446 2.514 Percent female unemployment 4.887 3.834 Percent male unemployment 6.970 4.494 Percent females in poverty 5.465 4.827 Percent males in poverty 4.697 4.040 Percent of households on public a ssistance 5.192 4.871 4.496 4.865 Percent hispanic 1.591 1.463 2.359 2.302 Residential mobility 1.976 2.219 1.696 1.718 Officer rate per 1,000 1.796 1.845 2.098 2.056 South 1.720 1.629 2.246 2.063

PAGE 72

72 Table 3-4. Principal components analys is after varimax rotation 1990 (N=178). VARIABLES Female Model Male Model Percent below Poverty .881 .836 Percent Unemployed .888 .941 Percent of Households on Public Assistance .893 .917 Eigenvalue 2.574 2.611 % Variance Explained 85.81% 87.037% Note: only factor loadings greater than 0.5000 are reported

PAGE 73

73 Table 3-5. Principal co mponents analysis after va rimax rotation 2000 (N=178). VARIABLES Female Model Male Model Percent below Poverty .903 .856 Percent Unemployed .903 .921 Percent of Households on Public Assistance .832 .886 Eigenvalue 2.546 2.574 % Variance Explained 85.851 85.809% Note: only factor loadings greater than 0.5000 are reported

PAGE 74

74 CHAPTER 4 RESULTS The following chapter presents the results of the cross-sect ional analyses and the pooled cross-sectional time series analyses. Cross-Sectional Analysis Tables 4-1 and 4-2 present the analysis from the zero-inflated negative binomial regression equations for maleand female-perpe trated intimate partner homicide counts in 1990 and 2000. Table 4-3 summarizes the findings. So me interesting findings emerge regarding exposure reduction, backlash, and economic de privation and marginal ization on maleand female-perpetrated intimate partner homicide. The following section will begin by explaining the findings from 1990 for both the fe male-perpetrated intimate partner model and the maleperpetrated intimate partner model and then will end with the cross-sectional results from the 2000 models. Female-Perpetrated Intima te Partner Homicide 1990 After examining the female-perpetrated in timate partner homicide model for 1990 it is clear that the first hypothesis is not supported. The only exposure reduction measure that showed slight significance was the percent of females divorced in cities (p < .1). It was found that as the percent of females divorced in c ities increases, the counts of fe male-perpetrated intimate partner homicide increase as well. Specifically, a one standard deviation incr ease in the percent of females divorced results in a 13% increase in female-perpetrated intimate partner homicides {[exp (.029 x 4.24) = 1.130]}. However, this is not in the predicted direc tion. According to the exposure reduction perspective, re ducing the exposure between violent couples (i.e., an increase in female % divorced) should reduce the likeli hood of female-perpetrated intimate partner homicide. Surprisingly, none of the other exposure reduction indi cators, including the measures

PAGE 75

75 for domestic violence service availability, are significantly related to female-perpetrated intimate partner homicide. Results are consistent with the second hypothe sis. None of the backlash indicators are significantly associated with female-perpetr ated intimate partner homicide in 1990 when controlling for a number of other stru ctural factors. It was hypothesized that as females status in terms of income, employment, and educational a ttainment increased male-perpetrated intimate partner homicide would increase as well. Backlash indicators were not predicted to be related to female-perpetrated intimate partner homicide. Consistent with the third hypothesis, econom ic deprivation and marginalization is significantly related to female-perpetrated intimate partner homicide in the predicted direction. That is, the higher the female economic deprivation index in large cities, th e higher the counts of female-perpetrated intimate partner homicide. Specifically, a one standard deviation increase in the economic deprivation index results in a 40% increase in female-perpetrated intimate partner homicides {[exp (.028 x 11.97 = 1.398]}. Just as predicted, it was hypot hesized that economic deprivation and marginalization would have a si gnificant impact on female-perpetrated intimate partner homicide, such that increases in econom ic deprivation and marginalization would be related to higher female-perpetrated intimate partner homicide in cities. In addition, despite findings from previous research none of the control variables are significant in the model. Residential mobility, the officer rate, log of the percent Hispanic, or whether or not the city was located in the south ar e not significantly related to female-perpetrated intimate partner homicide in 1990. However, all of the coefficien t signs are in the predicted direction.

PAGE 76

76Male-Perpetrated Intimate Partner Homicide 1990 After examining the male-perpetrated intimate partner homicide model for 1990 it is clear that contrary to the fi rst hypothesis, a number of the exposure reduction measures are related to male-perpetrated intimate partne r homicide. It was hypothe sized that exposure reduction indicators would impact female-perpetrated intimate partner homicide by reducing the contact between violent partners. However, th ree exposure reduction indicators are associated with male-perpetrated intimate partner homicid e. The number of legal services per 100,000 females is negatively associated with male-perpe trated intimate partner homicide. Specifically, a one standard deviation increase in the number of legal servi ces per 100,000 females results in a 15% decrease in male-perpetrated intimate part ner homicide counts {[exp (-.578 x .289 = .846]}. Moreover, the percent of males divorced is positively related to male-perpetrated intimate partner homicide. That is, a one standard deviat ion increase in the pe rcent of males divorced results in an 18% increase in male-perpetrated intimate partner homicide counts {[exp (.053 x 3.12 = 1.179]}. This contradicts the ideas behind exposure reduction. One would assume that as the exposure between partners decreased, particularly through divorce, in timate partner homicide would decrease as well. Furthermore, the number of shelters per 100, 000 females is positively related to maleperpetrated intimate partner homicide. That is, a one standard deviation incr ease in the number of shelters per 100,000 females results in a 26% in crease in male-perpetr ated intimate partner homicide counts {[exp (.828 x .276 = 1.257]}. No ne of the other exposure reduction indicators reached statistical significance. Also surprisingly, there is a lack of support for the second hypothesis. It was hypothesized that the backlash i ndicators would be significantly related to male-perpetrated intimate partner homicide, such that an increase in females status w ould be related to an

PAGE 77

77 increase in male-perpetrated intimate partner homicide as males attempt to gain control back in their relationships. However, none of the backlash indicators are signific antly related to maleperpetrated intimate partner homicide. Consistent with the third hypothesis, economic deprivation is significantly related to male-perpetrated intimate partner homicide counts in cities for 1990. The higher the male economic deprivation index in cities, the higher the counts of male-perpetrated intimate partner homicide. Specifically, a one standa rd deviation increase in the male economic deprivation index results in a 20% increase in male-perpetrated intimate partner homicide counts {[exp (.016 x 11.28 = 1.198]}. This is not surprising and is in the predicted direction. The only control variable that is significant in the model is whether or not the city was located in the South. Acco rding to the results, if the city is located in the South the likelihood of male-perpetrated intimate partner homicide incr eases. Other research has found similar results. For instance, DeWees and Parker (2003) found that female victimization is higher in the South. In summary, there are clear differences between which structural variables account for maleand female perpetrated intimate partner homicide in 1990. Exposure reduction measures influence both maleand female-perpetrated in timate partner homicides. However more of the measures influence male-perpetrated intimate part ner homicide and in particular, a couple of the domestic violence service indicators. None of th e backlash measures influence maleor femaleperpetrated intimate partner homicide in 1990. As predicted, economic deprivation and marginalization influence both maleand female-perpetrated intimate partner homicide in 1990. Female-Perpetrated Intima te Partner Homicide 2000 Examining the female-perpetrated intimat e partner homicide model for 2000 reveals some very different results than the 1990 m odel. In 2000, two of the exposure reduction indicators are significantly related to female-perpetrated intimate partner homicide. Inconsistent

PAGE 78

78 with the first hypothesis, the number of shelters per 100,000 females in cities is positively related to female-perpetrated intimate partner homicide. That is, a one standard deviation increase in the number of shelters per 100,000 fe males results in a 55% increase in female-perpetrated intimate partner homicide counts {[exp (.48 x .909 = 1. 547]}. Although a somewhat surprising finding, this was also found in the male-perpetrated intimate partner homicide model for 1990. In the predicted direction, the number of le gal services per 100,000 females in cities is negatively related to female-perpetrated intimat e partner homicide. That is, a one standard deviation increase in the number of legal services per 100,000 fe males results in a 28% decrease in female-perpetrated intimate partner homicid e counts {[exp (-.279 x 1.167 = .722]}. This was also seen in the male-perpetrated model for 1990. Contrary to the second hypothe sis, all three backlash indi cators are significantly related to female-perpetrated intimate partner homicide. Results suggest the higher the ratio of male to female median income (i.e., females status relativ e to males decreases) in cities, the lower the female-perpetrated intimate partner homicide. Sp ecifically, a one standard deviation increase in the ratio of male to female median income re sults in a 19% decrease in female-perpetrated intimate partner homicide counts {[exp (-1.156 x .180 = .812]}. Also, the ratio of male to female employment is negatively related to female-perpe trated intimate partner homicide. That is, the higher the ratio of male to female employment (i.e., gap widens) in cities, the lower femaleperpetrated intimate partner homicide. Specifically, a one standard deviation increase in the ratio of male to female employment results in a 23% decrease in female-perpetrated intimate partner homicide counts {[exp (-3.081 x .086 = .767]}. However, in the opposite dire ction, the ratio of males to females educational attainment is positively related to female-perpetrated intimate partner homicide. That is, a one standard deviati on increase in the ratio of males to females

PAGE 79

79 educational attainment (i.e., gap widens) results in a 22% increase in female-perpetrated intimate partner homicide counts {[e xp (2.042 x .098 = 1.222}. Not supporting the third hypothesis, the economic deprivation index is not significant in the 2000 model. This is surprising consideri ng that the economic deprivation index was significant at the .001 le vel in the 1990 model (p < .001). However, examining descriptive statistics does show a significant decrease in the female economic deprivation index from 1990 to 2000. The lack of significance in the economic deprivation index in 2000 may be due to the differences in the U.S. economy in the 1980s an d 1990s. The U.S. experienced a recession in the late 1980s until the early 1990s. This was followed by a substantial period of economic growth throughout the 1990s. The only control variable that is significant in the female-perpetrated model is the officer rate. That is, the higher the officer rate pe r 1,000 persons, the lower the count of femaleperpetrated intimate partner homicide. This sugges ts that more police officers would be able to react quicker to domestic violence calls and ther efore prevent many intimate partner homicides. Male-Perpetrated Intimate Partner Homicide 2000 The main difference between the male-perpetrated intimate partner homicide model in 2000 compared with 1990 is that the economic deprivation index is not significant. This is surprising, considering that economic depriva tion was significant at the .001 level in 1990 (p<.001). As stated earlier, the lack of significance in the economic deprivation index in 2000 may be due to the differences in the U.S. economy in the 1980s and 1990s. However, all other results are similar. Of the exposure reduction indicators, the per cent of males that are divorced in cities is positively related to male-perpetrated intimate pa rtner homicide. That is, the higher the percent of males divorced in cities, the higher the count of male-perpetrated intimate partner homicide.

PAGE 80

80 Specifically, a one standard deviation increase in the percent of males divorced results in a 12% increase in male-perpetrated intimate partner ho micide counts {[exp (.058 x 2.03 = 1.124]}. This follows the findings from the 1990 male-perpetrated model. Also, the number of shelters per 100,000 female s is positively related to male-perpetrated intimate partner homicide. That is, the higher the number of sh elters per 100,000 females in cities, the higher the count of male-perpetrated intimate partner homicide. Specifically, a one standard deviation increase in the number of shelters per 100,000 females results in a 20% increase in male-perpetrated intimate partne r homicide counts {[exp (.202 x .909 = 1.202]}. Though contradictory to the expos ure reduction perspective, it is not all that surprising considering that the same finding showed up in the 1990 male-perpetrated model. Moreover, the number of legal services per 100,000 females is negatively related to maleperpetrated intimate partner homicide. Specifically, the higher the number of legal services per 100,000 females in cities, the lower the count of male-perpetrated intimate partner homicide. That is, a one standard deviation increase in the number of legal services per 100,000 females results in a 25% decrease in male-perpetrated intimate partner homicide counts {[exp (-.244 x 1.167 = .752]}. This too, was a significant finding in 1990. Again surprisingly, none of the backlash i ndicators are related to male-perpetrated intimate partner homicide in 2000. It appears that m easures of females stat us relative to males are not predictive of male-perpetrated intimate partner homicide. Also, none of the control indicators are significant predictors of male-perpetrated intimate partner homicide. In summary, many of the structural indicat ors influence maleand female-perpetrated intimate partner homicide. Exposure reduction measures influence both maleand femaleperpetrated intimate partner homicides, with two domestic violence service variables

PAGE 81

81 significantly related to both; howev er both are not in the predicted direction. Contrary to what was predicted, backlash indicators influence fe male-perpetrated intimate partner homicide in 2000, but are not related to male-perpetrated inti mate partner homicide in 2000. Also opposite of the predictions, economic deprivation and margin alization did not influence either maleor female-perpetrated intimate partner homicide. Pooled Cross-Sectional Time Series Analysis: Fixed Effect Estimation Table 4-4 and 4-5 displays the results from the pooled cross-sectional Ordinary Least Squares fixed-effects regression models. Table 4-6 summarizes the findings. Interestingly, results suggest that the theoretical constructs explain changes in female-perpetrated intimate partner homicide, but not male-perpetrated intim ate partner homicide. Specifically, changes in female-perpetrated intimate partner homicide co unts are associated with variables used in exposure reduction explanations, but not with indicat ors used in backlash explanations. That is, cities that are characterized by increases in th e percent of females that are divorced are also characterized by decreasing counts of female -perpetrated intimate partner homicide. Surprisingly, the only domestic violence service variable that is significant in the femaleperpetrated intimate partner homicide model is the number of shelters per 100,000 females and the relationship between the cha nge in this variable and the change in female-perpetrated intimate partner homicide is in the opposite directi on of what was originally predicted. Cities that show increases in the number of shelters per 100,000 females al so show increases in femaleperpetrated intimate partner homicide counts. No significant relations hip is found between changes in the percent of unmarried partner households, the number of legal services available to females, the number of batterers counseling programs available to males, or the number of referral services offered to females and the fe male-perpetrated intimate partner homicide count.

PAGE 82

82 None of the changes in the indicators for backla sh are significantly related to the change in female-perpetrated intimate part ner homicide from 1990 to 2000. Furthermore, changes in the female economic deprivation index are significantly associated with changes in female-perpetrated intimate partner homicide counts. This suggests that intra-city changes in female economic depriv ation were strongly associated with within-city increases in female-perpetrated intimate partner homicide counts in the model. Cities that are characterized by increases in the percent of fema les in poverty, the percent of females that are unemployed, and the percent of households on public assistance from 1990 to 2000 are also characterized by increases in female-perpe trated intimate partner homicide counts. Overall, there is mixed support for the changes in the theoretical perspectives on the changes in female-perpetrated intimate partner ho micide counts from 1990 to 2000. A change in economic deprivation within cities is significantly associated with changes in female-perpetrated intimate partner homicide counts. Also, changes in two indicators of exposure reduction were related to changes in female-per petrated intimate partner homic ide counts; however one was in the opposite direction of what was predicted. Interestingly, support is not found for the ch anges in the variables identified by the theoretical perspectives on the changes in ma le-perpetrated intimate partner homicide counts from 1990 to 2000. Maleand female-perpetrated intimate partner homicides are unique events with different motives that requ ire a unique set of explanations. From this analysis, it appears that changes in factors that in fluence female-perpetrated intimat e partner homicide from 1990 to 2000 have no influence on the perpetration of ma le intimate partner hom icide from 1990 to 2000. Research examining these trends has acknowledg ed the greater support am ong certain factors in explaining female-perpetrated intimate partner ho micide. For instance, Dugan et al. (1999) find

PAGE 83

83 that indicators of exposure reduction primarily influence male intimate partner homicide victimization.

PAGE 84

84 Table 4-1. Zero-inflated negativ e binomial regression equations w ith coefficients (and Z-Scores) for gender-specific intimate partner homicide 1990. Female Model (N=178) Male Model (N=178) Exposure Reduction Percent divorced (gender-specific) .029 (1.84)+ .053 (3.26)** Percent unmarried households -.024 (-.62) -.024 (-.82) Male batterers programs ra te -.017 (-.27) .028 (.59) Shelter rate -.015 (-.04) .828 (2.72)** Legal service rate -.001 (-.00) -.578 (-1.99)* Referral services rate -.060 (-.20) -.202 (-.80) Backlash Ratio of male to female education .396 (-.75) -.221 (-.51) Ratio of male to female income .969 (1.39) -.041 (-.10) Ratio of male to female employment -.514 (-.83) -.256 (-.46) Economic Deprivation Economic deprivation index (gender-sp ecific) .028 (3.92)** .016 (2.80)** Controls Percent Hispanic (log) -.057 (-1.23) -.041 (-1.06) Residential mobility .010 (.88) -.002 (-.23) Officer rate -.069 (-.75) -.050 (-.68) South .255 (2.02) .332 (3.13)** Constant -12.44** -10.359** Maximum likelihood R-square .831 .821 Log likelihood -350.368** -462.997** + p < .10 p < .05 ** p < .01

PAGE 85

85 Table 4-2. Zero-inflated negativ e binomial regression equations w ith coefficients (and Z-Scores) for gender-specific intimate partner homicide 2000. Female Model (N=178) Male Model (N=178) Exposure Reduction Percent divorced (gender-specific) .052 (1.18) .058 (1.93)+ Percent unmarried households .105 (.79) .112 (1.35) Male batterers program rate .077 (.61) .076 (.94) Shelter rate .48 (3.95)** .202 (2.69)** Legal service programs rate -.279 (-2.41)* -.244 (-3.32)** Referral services rate .023 (.04) -.565 (-1.55) Backlash Ratio of male to female education 2.042 (1.72)+ .537 (.70) Ratio of male to female income -1.156 (-1.66)+ .019 (.05) Ratio of male to female employment -3.081 (-1.97)* -.040 (-.04) Economic Deprivation Economic deprivation index (gender-specific) .013 (.72) .009 (.71) Controls Percent Hispanic (log) .001 (.01) -.045 (-.71) Residential mobility -.020 (-.98) -.010 (-.84) Officer rate -.193 (-1.86)+ .057 (.95) South .218 (.91) .085 (.59) Constant -8.61** -11.99** Maximum likelihood R-square .783 .8084 Log likelihood -245.044** -424.747** + p < .10 p < .05 ** p < .01

PAGE 86

86 Table 4-3. Summary of cross-sectional results, 1990 and 2000. 1990 2000 Female Model Male Model Female Model Male Model Exposure Reduction % divorced (+) % divorced (+) # shelters (+) # legal services (-) # shelters (+) # legal services (-) % divorced (+) # shelters (+) # legal services (-) Backlash NONE NONE M/F income (-) M/F education (+) M/F employ (-) NONE Economic Deprivation EcoDep Index (+) EcoDep Index (+) NONE NONE Control NONE South (+) Officer Rate (-) NONE

PAGE 87

87 Table 4-4. Fixed-effects regressi on coefficients for the relationshi p between changes in predictor variables and changes in logged female intimate partner hom icide counts for 143a US cities, 1990-2000 b t Exposure Reduction Female percent divorced -.036+ -1.74 Percent unmarried households .065 1.26 Male batterers program rate .055 .60 Shelter rate .261* 2.15 Legal service programs rate -.035 -.40 Referral services rate -.268 -.90 Backlash Ratio of male to female education .095 .13 Ratio of male to female income -.184 -.29 Ratio of male to female employment .299 .32 Economic Deprivation Female economic deprivation index .035* 2.37 Controls Percent Hispanic (log) -.260+ -1.70 Residential mobility -.023 -1.03 Officer rate -.111 -.99 1990b -.030 -.10 Female dummy log zeroc -.754** -5.51 Female population 15 and over (log)d -.534** -2.81 Constant 8.90** 2.65 Within R2 .537 Hausman test .000** Notes: (a) 70 observations (35 groups) were dropped due to all zero outcomes (i.e. zero count for intimate partner homicide for both 1990 and 2000); (b) coefficients for pe riod effects are relative to the reference category, 2000; (c) represents a dummy variable for imputation of zero for log of a zero count; (d) female population is multiplied by a factor of three to account for the dependent variable multiplied by a factor of three. + p < .10 p < .05 ** p < .01

PAGE 88

88 Table 4-5. Fixed-effects regressi on coefficients for the relationshi p between changes in predictor variables and changes in logged male intimate partner homicide counts for 174a US cities, 19902000 b t Exposure Reduction Male percent divorced -.009 -.28 Percent unmarried households .034 .66 Male batterers program rate .113 1.40 Shelter rate .035 .33 Legal service programs rate .097 1.16 Referral services rate -.025 -.09 Backlash Ratio of male to female education .389 .78 Ratio of male to female income .360 .69 Ratio of male to female employment -.834 -.99 Economic Deprivation Male economic deprivation index .008 .66 Controls Percent Hispanic (log) -.013 -.09 Residential mobility -.010 -.77 Officer Rate .016 .17 1990b .338 1.16 Male dummy log zeroc -1.01** -6.79 Male population 15 and over (log)d .296 .68 Constant -1.99 -.34 Within R2 .303 Hausman Test .008** Notes: (a) 8 observations (4 groups) were dropped due to all zero outcomes (i.e. zero count for intimate partner homicide for both 1990 and 2000); (b) coefficients for pe riod effects are relative to the reference category, 2000; (c) represents a dummy variable for imputation of zero for log of a zero count; (d) male population is multiplied by a factor of 3 to account for the dependent variable multiplied by a factor of 3. + p < .10 p < .05 ** p < .01

PAGE 89

89 Table 4-6. Summary of change model results, 1990 to 2000. Female Model Male Model Exposure Reduction % divorced (-) # shelters (+) NONE Backlash NONE NONE Economic Deprivation EcoDep Index (+) NONE Control % Hispanic (-) NONE

PAGE 90

90 CHAPTER 5 DISCUSSION AND CONCLUSIONS The goal of t his study was to examine the trends in both maleand female-perpetrated intimate partner homicides from 1990 to 2000 to obtain a better understanding of the differential declines in maleand female-perpe trated intimate partne r homicide. Clearly, the decline in intimate partner homicide over time is a very complex issu e. This research suggests that one explanation will not suffice. This research was based on the theories of e xposure reduction, backlash, and economic deprivation and marginalization. Results suggest a lack of supp ort for these theories. Cross-sectional analyses are presented to evaluate the ability of the exposur e reduction perspective, the backlash perspective, and ideas behind economi c deprivation and marginalization in explaining maleand female-perpetrat ed intimate partner homicide counts in large cities at two time periods (1990 and 2000). Pooled cross-sectional time se ries analyses are presented to evaluate the effectiveness of the exposure reduction perspective, the backlash perspectiv e, and economic deprivation and marginalization in explaining the ch anges in maleand female -perpetrated intimate partner homicide in large cities over time. Discussion Dramatic changes were occurring in domesticity the status of female s in terms of income, employment and educational atta inment, economic deprivation, and domestic violence resources from 1990 to 2000 in large cities. For instance, during this time period the percentage of the population divorced decreased for both males and females, while the percentage of unmarried households increased. Furthermore, females have made gains in educa tional attainment and employment relative to ma les (i.e., status of females increased ). However, this is not true for females median income. Females median income relative to males actuall y decreased during this time period (i.e., status of females decreased). Also, economic deprivation in terms of poverty,

PAGE 91

91 unemployment, and public assistan ce income decreased for both ma les and females. Moreover, the availability of shelters and lega l services for females increased, whereas the availability of male batterers counseling programs and referral services decreased. Any of thes e changes may have an impact on the gender-specific trends in intimate partner homicide that ha ve been witnessed over the last couple of decades. Exposure Reduction Some of the structural predictors are relate d to intimate partner homicide perpetration. According to the exposure reduction perspective, it was hypothesized that measures that reduced the contact between violent partners would be asso ciated with female inti mate partner homicide perpetration in both 1990 and 2000. For 1990, cr oss-sectional analyses show that the exposure reduction measure of female percent divorced is in f act associated with female-perpetrated intimate partner homicide; however the relationship is in the opposite direction of what was predicted. Instead of the percent divorced being associated with a decrease in fema le-perpetrated intimate partner homicide, as the exposure reduction perspe ctive would predict, the percent of females divorced is associated with an increase in female-perpetrated in timate partner homicide. Other research has found similar results (Dugan et al ., 1999; 2003). One possibl e explanation may have to do with the dynamics that co me with the ending of a viol ent relationship. Research has suggested that the most dangerous time in a violen t relationship is when th e relationship is coming to an end (Campbell, 1992; Goetting, 1995). So me females may be reacting to male partners violence over the end of the relationship and may resort to lethal means to protect themselves. None of the other exposure reductio n measures were related to fema le-perpetrated intimate partner homicide in 1990. In 2000, two of the domestic violence serv ice variables reach si gnificance for femaleperpetrated intimate partner homicide, however in opposite directions. The number of shelters per

PAGE 92

92 100,000 females is associated with an increase in female-perpetrated intimate partner homicide. One reason for this finding may be that although the shelter rate is accounting for the number of shelters available per 100,000 fema les in cities, it may not be capturing shelter usage by females in abusive relationships. It is possible that the availability of shelters across cities is increasing but females are not taking adva ntage of them to reduce the exposure between themselves and their abusive partner. On the opposite end of the sp ectrum, females may in fact be using shelters to reduce the exposure between themselves and th eir violent partners, but once shelter stays end may find themselves in contact with their violen t partners and may have to resort to lethal violence to protect themselves, if adequate protection is not available. Furthe rmore, research has shown that the most dangerous time in an abusiv e relationship is when a partner is trying to end the relationship (Goetting, 1995). Some females ma y be reacting to male partners violence over the threat of the end of the rela tionship and may resort to lethal means to protect themselves. The second domestic violence service variab le associated with female-perpetrated intimate partner homicide is the number of legal services per 100,000 female s. This variable is associated with a decrease in female-perpetrated intimate partner homicide. It seems that as females are provided with more opportunities fo r legal assistance, such as in obtaining restraining orders, court accompaniment, legal cl inics or advocacy the opportunities for females to kill their male partners decreases. Interestingly, three of the six exposure re duction measures are a ssociated with maleperpetrated intimate partner homicides in 1990 a nd 2000. Similar to female -perpetrated intimate partner homicide in 2000, the percent of males divo rced is positively relate d to male-perpetrated intimate partner homicide. Based on suggestions fr om prior research (Goetting, 1995), it seems

PAGE 93

93 plausible that the end of a relationship ma y cause some males to use more violence towards their female partners, resulting in more male-perpetrated intimate partner homicide. The shelter rate and the legal se rvice rate were also significant ly related to male-perpetrated intimate partner homicide, however in opposite directions. Instead of reducing intimate partner homicide due to a decrease in opportunity, as the exposure reduction perspective would predict, the number of shelters per 100,000 females increases male intimate partner homicide perpetration. Although surprising, this finding is plausible in light of the rese arch that has found that maleperpetrated homicide is likely to occur when a partner is trying to leave an abusive relationship (Goetting, 1995). Some males may f eel threatened that their partne rs are leaving and staying at a shelter and may react in a lethal manner. More over, adequate protecti on may not be provided when females are ending th eir stay at a shelter. In addition, the number of legal services ava ilable to females is re lated to a decrease in male intimate partner homicide perpetration. It seems reasonable that, as females are provided with more opportunities for legal assistance, the opportunities fo r males to kill their female partners would decrease. Dugan et al. (2003) found similar results. That is they found that legal advocacy, measured as an index of the number of agencies with a separate budget for legal advocacy and the number of agencies that have lawy ers on staff, was related to fewer killings of white married females by their partners. Backlash According to the backlash perspective, it was hypothesized that measur es that may threaten male control and dominance would be associated with male intima te partner homicide perpetration in 1990 and 2000. Interestingly, support was not found for this hypothesis. Although surprising, backlash explanations may be less relevant in explaining male-perpetrated intimate partner homicide, because gender roles have changed over the years. Today, it is more acceptable for

PAGE 94

94 women to have jobs outside of the home and pursue higher education. Due to this transformation in gender roles, females economic and social stat us has increased. Furthermore, Felson (2006) surveyed 10,000 men and women about their spouse s behavior and concluded that males are no more controlling than females. Contrary to what was expected backlash correlates are associ ated with female-perpetrated intimate partner homicide in 2000. It appear s that, as the gap betw een male a nd female employment and income widens (i.e., females status relativ e to males decreases), femaleperpetrated intimate partner homicide decreases. Ho wever, as the gap between males and females educational attainment widens, female-perpetr ated intimate partner homicide increases. Females may react to disparities in income, employment and educational attainment just as it is predicted that males do, just in terms of frustration and not in terms of maintain ing control and dominance in the relationship. Along these lines, females appear to be taking their frustrations about the widening gap in educational attain ment out on their male partners, but not their frustrations about the widening gap in income or employment. Ande rson (1997) suggests that it is possible that females with lower status relative to males are more likely to engage in vi olence because they are less likely to be able to leave a violent relationship. That being said, females with a lower educational attainment status relative to male s may be less likely to be able to remove themselves from a violent situation, possibly be cause education affords females opportunities to learn about services available to reduce contact to prevent vi olence. Anderson (1997) found that women with slightly less education than their male partners have almost 3 times greater odds of perpetrating domestic violence th an women partnered with men processing the same amount of education (664).

PAGE 95

95Economic Deprivation and Marginalization According to the ideas behind economic deprivation and marginalization, it was hypothesized that measures of pove rty, unemployment and the percen t of the population on public assistance would be associated with both maleand female-perpetrated intimate partner homicides. Support for this hypothesis was found in 1990. Poverty, unemployment, and having to rely on public assistance income may significantly impact the dynamics within the home, either in terms of frustration or strain. Males and females may r eact to this strain by lashing out in a lethal manner at their intimate partners. Summary The results of this analysis offer support fo r economic deprivation an d marginalization in explaining both maleand female-perpetrated in timate partner homicide in 1990. Results also suggest an association between some of the variables representing the exposure reduction perspective and maleand female-perpetrated inti mate partner homicide in both 1990 and 2000. In general, the three exposure reduction indicators that are consistently signifi cant in 1990 and 2000 are gender-specific percent divorced and two indi cators measuring the av ailability of domestic violence services the shelter rate and the legal service rate. Importan tly, these measures are related to both maleand female-p erpetrated intimate partner homicides in a similar fashion. A one standard deviation increase in any of these measur es results in a significan t percentage change in intimate partner homicide counts. The availability of legal services in c ities supports the exposure reduction hypotheses. That is, an increase in the number of legal services per 100,000 females is related to an increase in male-p erpetrated intimate pa rtner homicide in 19 90 and 2000 and femaleperpetrated intimate partner homicide in 2000. However, not all of the significant exposure reduction measures influence intimate partner homicide in the predicted directi on. For instance, an increase in th e percent divorced increases both

PAGE 96

96 male and female intimate partner homicide. Furthermore, results suggest that an increase in the number of shelters per 100,000 females is related to an incr ease in male-perpetrated intimate partner homicide in 1990 and 2000 and female-perpetrated intimate partner homicide in 2000. This is a main domestic violence servic e measure that has historically been used to measure domestic violence services and was expect ed to have an exposure reduction effect on intimate partner homicide perpetration. Instead, it appears that th is measure is associated with more males and females reacting out in a lethal manner towards their intimate part ner, whether it is for protection or to maintain control and domina nce. Dugan et al. (2003) also f ound a backlash effect for one of their measures of domestic viol ence services. Specifi cally, they found that prosecutors willingness to prosecute violators of protection orders was related to an incr ease in homicides of married and unmarried white females and unmarried African-A merican males. They concluded that, although prosecutors willingness to prosecute is advantageous, if this willi ngness comes with out also being able to provide adequate protection for victims, homicide may result. As st ated earlier, if these services arent being utilized by females, they will be less likely to ha ve an exposure reduction effect on intimate partner homicide Also, given the fact that the likelihood of intimate partner homicide is extremely high when the relationship is ending, domestic viol ence resources need to provide adequate protection when th e relationship is ending or all th e good intentions of increasing shelter availability will not prevent intimate partner violence and homicide. Indicators measuring the backlash perspectiv e do not predict male-p erpetrated intimate partner homicide in 1990 or 2000 as was hypothesized. Interestingly, backlash indicators do reach significance in the 2000 female-per petrated intimate partner homicide model. However, the relationship between these variables and female-perpe trated intimate partner homicide is complex.

PAGE 97

97 A one standard deviation increase in any of thes e three ratio measures re sults in a significant percentage change in the female-perpetrated inti mate partner homicide in 2000. An increase in the ratio of median income and the ratio of male to female empl oyment has a negative effect on female-perpetrated intimate homici de (i.e., decreases), while an in crease in the ratio of male to female education has a positive ef fect (i.e., increases) on female-perpetrated intimate partner homicide. Changes Over Time In regards to changes over time, changes in structural variables fr om 1990 to 2000 were related to intimate partner homicide perpetration ov er time, however this is only true for femaleperpetrated intimate partner homicide. Interestingl y, although many of the structural predictors influence male-perpetrated intima te partner homicide cross-sectionally, these perspectives did not explain male-perpetrated intimate partner homici de over time at all. Results suggest that c ities that are characterized by changes in the percent of females divorced from 1990 to 2000 are characterized by changes in female-perpetrated intimate partner homicide counts. Also, cities that show changes in the numb er of shelters per 100,000 fema les from 1990 to 2000 also show changes in female-perpetrated intimate partner homicide counts. Furthe rmore, cities that are characterized by changes in the pe rcent of females in poverty, the percent of females that are unemployed, and the percent of households on public assistance from 1990 to 2000 are also characterized by changes in female-perpetrated intimate partner homicide counts. Lastly, cities that are characterized by cha nges in the logged percent Hisp anic population from 1990 to 2000 are characterized by changes in female-perpe trated intimate partner homicide counts. Overall, there is mixed support for the changes in the theoretical perspectives on the changes in female-perpetrated intimate pa rtner homicide counts from 1990 to 2000 and absolutely no support on the changes in male-p erpetrated intimate partner homicide counts.

PAGE 98

98 Changes in economic deprivation within cities are significantly associated with changes in female-perpetrated intimate partner homicide counts. Also, changes in two indicators of exposure reduction were related to changes in female-perpetrated intimate partner homicide counts; however the shelter rate was in the opposite direction of wh at was predicted and increased shelter availab ility appears to be associated with more harm than good if shelters are not being taken advantage of or if she lters are not providing adequate protection. Limitations There are some limitations to this research. First and foremost, the data utilized for the dependent variables was based on Supplemental Homici de Files. The deficiencies of the database have been well-documented (Junger-Tas and Ma rshall, 1999). These include underreporting, errors in the assignment of relationships, failure to identify relationship in a substantial portion of cases, and other missing information SHFs include information on homicides that were submitted by law enforcement agencies to the FBI. Submission of homicide information is completely voluntary and therefore not 100% accurate. In ad dition, many homicides are coded as having an unknown victim-offender relationshi p. A weighting proce dure was utilized to acco unt for this lack of reporting and lack of relationship information. Moreover, it is acknowledged that actual usag e of the domestic vi olence services was unable to be captured in cities. Ju st because shelters, legal servi ces, batterers counseling programs, and referrals are offered does not mean they are being utilized to the fullest by individuals in violent relationships. Also, the size of the domes tic violence service programs (i.e., service capacity) was unable to be captured. If domestic violence services are not being utilized or if programs have limited capacity, domestic violence services would have a li mited effect on intimate partner homicide.

PAGE 99

99 Also, it was difficult to divide the structural factors of inte rest into the three theoretical categories. It is possible that ma ny of the structural indicators could have fallen into another theoretical category. For instance, although for this study females status relative to males was considered to represent the backlash hypothesis, other research has considered females status relative to males as measuring exposure reduction, because an increase in the status of females affords them more opportunities to leave abusive relationships. In additi on, research has used females status relative to males to represent economic deprivation and marginalization. For the current study, all structural indicators were di vided into theoretical categories based on the developed hypotheses. Furthermore, data was limited to two time points (1990 and 2000) Although the 1990s witnessed drastic changes in inti mate partner homicide perpetration, changes in intimate partner homicide have been witnessed sin ce 1976. It would be interesting to expand the data to include earlier data. Also, data was limite d to large cities with a populati on over 100,000. It would be useful to examine data in a ll sizes of cities over time. Future Research Future research on intimate partner homicide could take many directions. First, it would be useful to examine intimate partner homicide as well as intimate partner aggravated assault, in addition to the access or availability of medical centers. Aggrav ated assault and homicide are similar events, with the exception of the outcome (i.e., one ends in death, whereas the other does not). Harris et al. (2002) argues that research would benefit by focusing on serious assaults, with only a small percentage that actually end in death. They show that the decl ine in the homicide rate parallels the increase in medical technology and support services. It is seems reasonable that the availability of medical care (i.e., trauma cent ers) may influence whether an intimate partner aggravated assault remained an aggravated assaul t or resulted in death and was categorized as an

PAGE 100

100 intimate partner homicide. Many intimate partner homicides may ha ve been aggravated assaults that became homicides when ade quate medical care was not available. Greater availability of medical resources may prevent many intimate part ner homicides and thus is an important factor and should be considered when examining intimate partner homicide over time. Also, it would be important to examine the cities that were excluded from the current analysis due to lack of reporting. The majority of the cities exclude d were located in the state of Florida. Research has shown that homicide rates are higher in Sout hern states (DeWees and Parker, 2003; Smith and Brewer, 1992). If intimate partner ho micide data could be gathered, it would be interesting to examine intimate pa rtner homicide in large cities in Florida to see if results were similar with the results fr om the current study. Furthermore, it would be important to include ot her factors in the analysis especially the age structure of the cities. Browne et al. (1999) examined age-specific ra tes of intimate partner homicide from 1980 to 1 995 and found that different trends emerge for different age groups and for males and females. They found that a decline in female-perpetr ated intimate partner homicide occurred in all age groups, except the age group 13 to 17 years old (rates ha ve remained low). The greatest decreases occurred in the 25 to 34 an d the 35 to 44 year ol d age groups. For maleperpetrated intimate partner homicide, increases were found for the 13 to 17 and the 18 to 24 year old age groups, while decreases were seen in the 25 to 34 and the 35 to 44 year old age groups. Since differential trends in intima te partner homicide have been witnessed for various age groups and younger age groups are more likely to participate in inti mate partner homicide, it would be important to control for ag e in future analyses. Conclusion This research offers many cont ributions to the literature on in timate partner homicide. Most importantly, this research examin es the dynamic nature of a nu mber of different structural

PAGE 101

101 variables on both maleand female-perpetrated intimate partner homici de over time. This is something that has not been done in the literature to date. The tim e period covered includes a time in history where dramatic change s were occurring in domesticity, the status of fe males, economic deprivation and marginalization, and the availab ility of domestic violence resources with the enactment of the Violence of Against Women Act in 1994. A main finding of this resear ch speaks to the ineffectiveness of domestic violence services. The number of shelters per 100,000 females is significantly related to intimate partner homicide in 1990 and 2000 and over time. Despit e all the efforts to increase she lter availability to females in violent relationships, it appears that the increase in availability is actually associated with an increase in intimate part ner homicide. For instance, in 1990 a nd 2000 the increase in the shelter rate was related to an increase in male-perpetrated intimate partner homicide. Furthermore, not only was this finding seen in 200 0 for female-perpetrated intimate partner homicide, this finding was witnessed over time as well. Efforts to prevent domestic viol ence and homicide need to make sure that they also provide adequate protection during times that are characterized by increased violence. This research adds to the knowledge of th e decline in female-perpetrated intimate homicide, but still leaves explanations behind the decline in male-perpetr ated unclear. Research needs to continue to examine ge nder-specific intimate partner ho micide, but particularly maleperpetrated intimate partner homicide. Surprisi ngly, none of the three theoretical perspectives were associated with male-perpetrated intimate partner homicide over ti me, whereas some of the predictors were associated with female-perpetrat ed intimate partner homicide over time. It is extremely important to gain a better understanding of predictors of male-perpetrated intimate partner homicide, given the fact that male-perpetrated intimate partner homicide has not decreased

PAGE 102

102 as sharply as female-perpetrated intimate part ner homicide over time de spite the increase in domestic violence resources geared towards preventi ng violence against females. It is essential to determine what really has influe nced male-perpetrated intimate partner homicide over time. This may be the only way to for us to see a larger decline in these types of homicides.

PAGE 103

103 APPENDIX A GLOSS ARY Place Places, for the reporting of decennial cen sus data, include census designated places, consolidated cities, and incorporated places. South Region South Atlantic Division: Delaware, Maryland, District of Columbia, Vi rginia, West Virginia, North Carolina, South Carolina, Georgia, Florida East South Central Division: Kentucky, Tennessee, Alabama, Mississippi West South Central Division: Arkansas, Louisiana, Oklahoma, Texas Employed All civilians 16 years old a nd over who were either (1) at work those who did any work at all during the reference week as pa id employees, worked in their own business or profession, worked on their own farm, or wo rked 15 hours or more as unpaid workers on a family farm or in a family business; or (2) we re with a job but not at work those who did not work during the reference w eek, but who had jobs or busin esses from which they were temporarily absent because of illnes s, bad weather, industrial dispute, vacation, or other personal reasons. Excluded from the employed are people whose only activity consisted of work around their own house (painting, repa iring, or own home housework) or unpaid volunteer work for religious, charitable, and simila r organizations. Also excluded are all institutionalized people and people on active duty in the United States Armed Forces. Unemployed All civilians 16 years old and over were cl assified as unemployed if they were neither at work nor with a job but not at work during the reference week, were looking for

PAGE 104

104 work during the last 4 weeks, and were availabl e to start a job. Also included as unemployed were civilians 16 years old and ove r who: did not work at all dur ing the reference week, were on temporary layoff from a job, had been informed th at they would be recalled to work within the next 6 months or had been given a date to return to work, and were available to return to work during the reference week, except for temporary illness. Civilian labor force Consists of people classified as employed or unemployed in accordance with the criteria described above. Labor force All people classified in the civili an labor force (i.e., employed and unemployed people), plus members of the U.S. Armed Forces (people on active duty with the United States Army, Air Force, Navy, Marine Corps, or Coast Guard). Not in labor force All people 16 years old and over who are not classified as members of the labor force. This category consists mainly of st udents, individuals taking care of home or family, retired workers, seasonal workers enumerated in an off-season who were not looking for work, institutionalized people (all institutionalized people are placed in this cat egory regardless of any work activities they may have done in the refe rence week), and people doing only incidental unpaid family work (fewer than 15 hours during the reference week). Household A household includes all of the people who occupy a housing unit. (People not living in households are classifi ed as living in group quarters. ) A housing unit is a house, an apartment, a mobile home, a group of rooms, or a single room occupied (or if vacant, intended for occupancy) as separate living quarters Householder The data on relationship to householder were derived from the question, How is this person related to Person 1, which was aske d of Persons 2 and higher in housing units. One person in each household is designated as the householder (Person 1). In most cases, the

PAGE 105

105 householder is the person, or one of the people, in whose name the home is owned, being bought, or rented. If there is no such person in the household, any adult household member 15 years old and over could be designated as the householder (i.e., Person 1) Households are classified by type according to the sex of the householder a nd the presence of rela tives. Two types of householders are distinguished: family house holders and nonfamily householders. A family householder is a householder living with one or more individuals re lated to him or her by birth, marriage, or adoption. The householder and all of the people in the house hold related to him or her are family members. A nonfamily househol der is a householder liv ing alone or with nonrelatives only. Unmarried partner An unmarried partner is a person who is not related to the householder, who shares living quarters, and who has a clos e personal relationship with the householder. Unmarried-Partner Household An unmarried-partner household is a household that includes a householder and an unmarried pa rtner. An unmarried partner can be of the same or of the opposite sex of the householder. An unma rried partner in an unmarried-partner household is an adult who is unrelated to the householder, but shares living quarters and has a close personal relationship with the householder. An unmarried-par tner household may also be a family household or a nonfamily household, depending on the presence or absence of another person in the household who is related to the householder. Th ere may be only one unmarriedpartner per household, and an unmarried part ner may not be included in a married-couple household as the householder cannot have bot h a spouse and an unmarried partner. Income of individuals Income for individuals is obtaine d by summing the eight types of income for each person 15 years old and over. Th e characteristics of individuals are based on the time of enumeration (April 1, 2000), even t hough the amounts are for calendar year 1999.

PAGE 106

106Median income The median divides the income distribut ion into two equal parts: one-half of the cases falling below the median income and one-half above the median. For households and families, the median income is based on the di stribution of the total number of households and families including those with no income. The me dian income for individuals is based on individuals 15 years old and over with income. Median income for households, families, and individuals is computed on the ba sis of a standard distribution (s ee the Standard Distributions section under Derived Measures ). Median income is rounded to the nearest whole dollar. Median income figures are calculated using linear interpolation if the width of the interval containing the estimate is $2,500 or less. If the wi dth of the interval containing the estimate is greater than $2,500, Pareto in terpolation is used. (For more information on medians and interpolation, see Derived Measures.) Divorced -This category includes people who are lega lly divorced and who have not remarried. How Poverty Status is Determined The poverty status of families and unrelated individuals in 1999 was determined using 48 thresholds (income cutoffs) arranged in a two dimensional matrix. The matrix consists of family size (from 1 person to 9 or more people) crossclassified by presence and number of family members under 18 years old (from no children presen t to 8 or more childre n present). Unrelated individuals and 2-person families were further differentiated by the age of the reference person (RP) (under 65 years old and 65 years old and over). To determine a persons poverty status, one compares the persons total family income with the poverty threshold appropriate for that persons family size and composition (see table below). If the total income of that persons family is less than the threshold appropriate for that family, then the person is considered poor, togeth er with every member of his or her family. If a person is not

PAGE 107

107 living with anyone related by birth, marriage, or adoption, then the persons own income is compared with his or her poverty threshold. Individuals for whom povert y status is determined Poverty status was determined for all people except institutionalized people, people in military group quarters, people in college dormitories, and unrelated individuals under 15 years old. These gr oups also were excluded from the numerator and denominator when calculating poverty rates. They are considered neither poor nor nonpoor. (Census Glossary) Residence 5 years ago The data on residence 5 years earlie r were derived from answers to long-form questionnaire Item 15, which was aske d of a sample of the population 5 years old and over. This question asked for the state (or foreig n country), U.S. county, city or town, and ZIP Code of residence on April 1, 1995, for those people who reported that on th at date they lived in a different house than their current residence. Residence 5 years earlier is used in conjunction with location of current residence to determin e the extent of residential mobility of the population and the resulting redistribution of the population across the various states, metropolitan areas, and regions of the country. Summary File 1 This Census 2000 file presents 100percent population and housing data for the total population, for 63 race categories, a nd for many other race and Hispanic or Latino categories. The data include age, sex, households, household relations hip, housing units, and tenure (whether the residence is ow ned or rented). Also included are selected characteristics for a limited number of race and Hispanic or Latino categories. The data are available for the U.S., census regions, census divisions, states and st atistically equivalent entities, counties and statistically equivalent entities, county subdi visions, places, census trac ts, block groups, census blocks, metropolitan areas, urban areas, Americ an Indian and Alaska Native areas, tribal

PAGE 108

108 subdivisions, Hawaiian home lands, Congressional district s, and ZIP Code ta bulation areas. Data are available down to the block level for many ta bulations, but only to th e census tract level for others. Available on CD-ROM, DVD, and American FactFinder. Summary File 3 This Census 2000 file presents da ta on population and housing long-form subjects, such as income and education. It incl udes population totals for an cestry groups. It also includes selected characteristics for a limited number of race and Hispanic or Latino categories. The data are available for the U.S., census regi ons, census divisions, st ates and statistically equivalent entities, counties and statistically equivalent entities, county subdivisions, places, census tracts, block groups, metropolitan areas, urban areas, American Indian and Alaska Native areas, tribal subdivisions, Hawa iian home lands, Congressional districts, and ZIP Code tabulation areas. Available on CD-ROM DVD, and American FactFinder. UCR Uniform Crime Reports

PAGE 109

109 APPENDIX B CORRELATION MATRIX FOR EXPLA NATORY AND OUTCO ME VARIABLE 1990.

PAGE 110

110 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 1. Female-perp IPH 1.00 2. % Fem poverty .359 1.00 3. % Fem unempl .300 .782 1.00 4. % HH pub assis. .244 .786 .793 1.00 5. M/F income -.221 -.282 -.093 -.265 1.00 6. M/F education -.190 -.393 -.328 -.364 .429 1.00 7. M/F % employ -.003 .181 .249 .212 .220 .208 1.00 8. % fem div .005 -.106 -.124 -.041 -.076 .050 -.126 1.00 9. % unmarried .003 -.020 .053 .124 -.236 -.084 -.105 -.122 1.00 10. Shelter rate -.281 .024 -.107 -.026 .118 -.039 -.085 .036 -.129 1.00 11. Legal serv rate -.223 .093 .011 .026 .078 -.017 -.068 .079 -.086 .678 1.00 12. Male batt. rate -.078 .052 -.048 .016 -.077 -.178 -.071 .127 -.118 .293 .316 1.00 13. Referral rate -.115 -.048 .084 .034 .034 -.046 -.072 .025 .017 .164 -.009 .133 1.00 14. Res. mobility -.244 -.460 -.431 -.442 .046 .111 -.226 .213 .225 .028 .016 .009 .079 1.00 15. Hispanic(log) -.058 -.125 .110 .072 -.159 .032 .152 .109 .198 -.126 -.105 -.045 .077 .279 1.00 16. Officer rate .347 .467 .305 .388 -.403 -.345 -.146 -.132 -.004 -.014 -.044 .039 -.049 -.449 -.167 1.00 17. South .183 .189 .045 -.195 .014 -.083 -.041 -.189 -.234 -.008 .010 .074 -.067 -.025 -.262 .105 1.00 1. Male-perp IPH 1.00 2. % Male poverty .260 1.00 3. % Male unempl .199 .786 1.00 4. % HH pub assis. .195 .766 .863 1.00 5. M/F income -.250 -.322 -.193 -.265 1.00 6. M/F education -.177 -.405 -.335 -.364 .429 1.00 7. M/F % employ .006 .192 .218 .212 .220 .208 1.00 8. % Male div .050 .001 .024 .029 -.094 -.009 -.152 1.00 9. % Unmarried .004 -.021 .063 .124 -.236 -.084 -.105 -.107 1.00 10. Shelter rate -.280 .011 -.053 -.026 .118 -.039 -.085 .054 -.129 1.00 11. Legal Serv rate -.245 .091 .050 .026 .078 -.017 -.068 .079 -.086 .678 1.00 12. Male Batt. rate -.039 .062 -.018 .016 -.077 -.178 -.071 .137 -.118 .293 .316 1.00 13. Referral rate -.121 -.033 .063 .034 .034 -.046 -.072 .005 .017 .164 -.009 .113 1.00 14. Res. mobility -.227 -.389 -.537 -.442 .046 .111 -.226 .138 .225 .028 .016 .009 .079 1.00 15. Hispanic(log) .018 -.041 .075 .072 -.159 .032 .152 .014 .198 -.126 -.105 -.045 .077 .279 1.00 16. Officer rate .323 .426 .444 .388 -.403 -.345 -.146 -.048 -.004 -.014 -.044 .039 -.049 -.449 -.167 1.00 17. South .122 .151 -.160 -.195 .014 -.083 -.041 -.188 -.234 -.008 .010 .074 -.067 -.025 -.262 .105 1.00

PAGE 111

111 APPENDIX C CORRELATION MATRIX FOR EXPLANAT ORY AND OUTCOME VARIABLES 2000.

PAGE 112

112 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 1. Female-perp IPH 1.00 2. % Fem poverty .257 1.00 3. % Fem unempl .219 .816 1.00 4. % HH pub assis. .147 .751 .751 1.00 5. M/F income -.245 -.256 -.200 -.178 1.00 6. M/F education -.148 -.365 -.365 -.326 .476 1.00 7. M/F % employ -.124 -.201 -.182 -.122 .396 .535 1.00 8. % fem div .032 -.061 -.146 -.145 -.009 -.038 -.245 1.00 9. % unmarried .067 .339 .382 .537 -.294 -.358 -.267 .209 1.00 10. Shelter rate -.224 .089 .095 .115 -.012 -.079 -.182 .010 .150 1.00 11. Legal serv rate -.197 .030 .033 .072 -.092 -.058 -.168 -.008 .124 .709 1.00 12. Male batt. rate -.100 .018 -.004 -.015 -.021 -.123 -.055 -.021 -.014 .347 .431 1.00 13. Referral rate -.137 .047 .065 -.002 .109 -.035 .065 -.034 -.060 .047 -.253 .086 1.00 14. Res. mobility -.084 -.164 -.155 -.251 -.289 .004 -.122 .162 .117 .008 .058 .019 -.013 1.00 15. Hispanic(log) -.058 -.061 .085 .177 -.053 .061 .477 -.345 .133 -.182 -.191 -.008 .108 .153 1.00 16. Officer rate .311 .482 .425 .353 -.424 -.224 -.372 -.011 .152 .004 .072 .000 -.103 -.210 -.282 1.00 17. South .096 .099 .001 -.293 -.019 -.023 -.078 .070 -.523 -.115 -.121 .112 .053 .118 -.213 .107 1.00 1. Male-perp IPH 1.00 2. % Male poverty .210 1.00 3. % Male unempl .126 .788 1.00 4. % HH pub assis. .151 .757 .816 1.00 5. M/F income -.226 -.323 -.218 -.178 1.00 6. M/F education -.105 -.370 -.405 -.326 .476 1.00 7. M/F % employ -.007 -.124 -.233 -.122 .396 .535 1.00 8. % Male div -.013 .000 .131 -.009 -.034 -.123 -.348 1.00 9. % Unmarried .048 .371 .484 .537 -.294 -.358 -.267 .302 1.00 10. Shelter rate -.327 .054 .130 .115 -.012 -.079 -.182 .136 .150 1.00 11. Legal serv rate -.241 .012 .062 .072 -.092 -.058 -.168 .085 .124 .709 1.00 12. Male batt. rate -.105 .003 -.010 -.015 -.021 -.123 -.055 .022 -.014 .347 .431 1.00 13. Referral rate -.143 .040 .066 -.002 .109 -.035 .065 .039 -.060 .047 -.253 .086 1.00 14. Res. mobility -.099 -.081 -.195 -.251 -.289 .004 -.122 .065 .117 .008 .058 .019 -.013 1.00 15. Hispanic(log) .071 .080 .012 .177 -.053 .061 .477 -.434 .133 -.182 -.191 -.008 .108 .153 1.00 16. Officer rate .338 .411 .435 .353 -.424 -.224 -.372 .125 .152 .004 .072 .000 -.103 -.210 -.282 1.00 17. South .005 .026 -.154 -.293 -.019 -.023 -.078 .040 -.523 -.115 -.121 .112 .053 .118 -.213 .107 1.00

PAGE 113

113 APPENDIX D SIMPLIFIED CORRELATION MATRIX FOR EXPLANATORY AND OUT COME VARIABLES 1990.

PAGE 114

114 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 1. Female-perp IPH 1.00 2. Econ. dep. index .359 1.00 3. M/F income -.221 -.241 1.00 4. M/F education -.190 -.393 .429 1.00 5. M/F % employment -.003 .209 .220 .208 1.00 6. % Fem divorced .005 -.116 -.076 .050 -.126 1.00 7. % Unmarried HH .003 .000 -.236 -.084 -.105 -.122 1.00 8. Shelter rate -.281 -.012 .118 -.039 -.085 .036 -.129 1.00 9. Legal service rate -.223 .074 .078 -.017 -.068 .079 -.086 .678 1.00 10. Male batt. rate -.078 .026 -.077 -. 178 -.071 .127 -.118 .293 .316 1.00 11. Referral rate -.115 -.012 .034 -.046 -.072 .025 .017 .164 -.009 .133 1.00 12. Res. mobility -.244 -.473 .046 .111 -.226 .213 .225 .028 .016 .009 .079 1.00 13. % Hispanic (log) -.058 -.063 -.159 .032 .152 .109 .198 -.126 -.105 -.045 .077 .279 1.00 14. Office rate .347 .442 -.403 -.345 -.146 -.132 -.004 -.014 -.044 .039 -.049 -.449 -.167 1.00 15. South .183 .156 .014 -.083 -.041 -.189 -.234 -.008 .010 .074 -.067 -.025 -.262 .105 1.00 1. Male-perp IPH 1.00 2. Econ. dep. index .254 1.00 3. M/F income -.250 -.298 1.00 4. M/F education -.177 -.403 .429 1.00 5. M/F % employment .006 .210 .220 .208 1.00 6. % Male divorced .050 .016 -.094 -.009 -.152 1.00 7. % Unmarried HH .004 .004 -.236 -.084 -.105 -.107 1.00 8. Shelter rate -.280 -.008 .118 -.039 -.085 .054 -.129 1.00 9. Legal Service rate -.245 .083 .078 -.017 -.068 .079 -.086 .678 1.00 10. Male batt. rate -.039 .040 -.077 -. 178 -.071 .137 -.118 .293 .316 1.00 11. Referral rate -.121 -.005 .034 -.046 -.072 .005 .017 .164 -.009 .133 1.00 12. Res. mobility -.227 -.454 .046 .111 -.226 .138 .225 .028 .016 .009 .079 1.00 13. % Hispanic (log) .018 -.007 -.159 .032 .152 .014 .198 .198 -.105 -.045 .077 .279 1.00 14. Office rate .323 .452 -.403 -.345 -.146 -.048 -.004 -.004 -.044 .039 -.049 -.449 -.167 1.00 15. South .122 .061 .014 -.083 -.041 -.188 -.234 -.34 .010 .074 -.067 -.025 -.262 .105 1.00

PAGE 115

115 APPENDIX E SIMPLIFIED CORRELATION MATRIX FOR EXPLANATORY AND OUT COME VARIABLES 2000.

PAGE 116

116 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 1. Female-Perp IPH 1.00 2. Econ. Dep. Index .229 1.00 3. M/F Income -.245 -.237 1.00 4. M/F Education -.148 -.380 .476 1.00 5. M/F % Employment -.124 -.184 .396 .535 1.00 6. % Fem Divorced .032 -.114 -.009 -.038 -.245 1.00 7. % Unmarried HH .067 .449 -.294 -.358 -.267 .209 1.00 8. Shelter Rate -.224 .107 -.012 -.079 -.182 .010 .150 1.00 9. Legal Service Rate -.197 .049 -.092 -.058 -.168 -.008 .124 .709 1.00 10. Male Batt. Rate -.100 .003 -.021 -.123 -.055 -.021 -.014 .347 .431 1.00 11. Referral Rate -.137 .036 .109 -.035 .065 -.034 -.060 .047 -.253 .086 1.00 12. Res. Mobility -.084 -.208 -.289 .004 -.122 .162 .117 .008 .058 .019 -.013 1.00 13. % Hispanic (log) -.058 .052 -.053 .061 .477 -.345 .133 -.182 -.191 -.008 .108 .153 1.00 14. Office Rate .311 .462 -.424 -.224 -.372 -.011 .152 .004 .072 .000 -.103 -.210 -.282 1.00 15. South .096 -.059 -.019 -.023 -.078 .070 -.523 -.115 -.121 .112 .053 .118 -.213 .107 1.00 1. Male-Perp IPH 1.00 2. Econ. Dep. Index .180 1.00 3. M/F Income -.226 -.260 1.00 4. M/F Education -.105 -.390 .476 1.00 5. M/F % Employment -.007 -.161 .396 .535 1.00 6. % Male Divorced -.013 .030 -.034 -.123 -.348 1.00 7. % Unmarried HH .048 .500 -.294 -.358 -.267 .302 1.00 8. Shelter Rate -.327 .104 -.012 -.079 -.182 .136 .150 1.00 9. Legal Service Rate -.241 .051 -.092 -.058 -.168 .085 .124 .709 1.00 10. Male Batt. Rate -.105 -.008 -.021 -.123 -.055 .022 -.014 .347 .431 1.00 11. Referral Rate -.143 .032 .109 -.035 065 .039 -.060 .047 -.253 .086 1.00 12. Res. Mobility -.099 -.189 -.289 .004 -.122 .065 .117 .008 .058 .019 -.013 1.00 13. % Hispanic (log) .071 .110 -.053 .061 .477 -.434 .133 -.182 -.191 -.008 .108 .153 1.00 14. Office Rate .338 .426 -.424 -.224 -.372 .125 .152 .004 .072 .000 -.103 -.210 -.282 1.00 15. South .005 -.154 -.019 -.023 -.078 .040 -.523 -.115 -.121 .112 .053 .118 -.213 .107 1.00

PAGE 117

117 APPENDIX F VARIANC E INFLATION FACTORS FOR SIMPLIFIED MODELS Fem-Perp90 Male-Perp90 Fem-Perp00 Male-Perp00 Female economic deprivation index 1.900 2.028 Male economic deprivation index 1.987 2.006 Ratio of male to female education 1.549 1.546 2.068 2.084 Ratio of male to female income 1.615 1.614 2.039 2.034 Ratio of male to female employment 1.438 1.457 2.369 2.381 Percent female divorced 1.226 1.450 Percent male divorced 1.197 1.623 Percent unmarried households 1.343 1.314 2.799 2.908 Male batterers programs per 100,000 males 1.226 1.227 1.397 1.397 Shelters per 100,000 females 2.056 2.054 2.391 2.385 Legal service programs per 100,000 females 2.056 2.061 2.864 2.855 Referral services per 100,000 females 1.108 1.111 1.290 1.308 Percent hispanic (log) 1.348 1.361 2.078 2.176 Residential mobility 1.908 1.890 1.634 1.616 Officer rate per 1,000 1.780 1.789 2.076 2.026 South 1.237 1.242 1.959 1.891

PAGE 118

118 APPENDIX G CORRELATION MATRICES FOR DOMESTIC VIOLE NCE SERVICE VARIABLES. 1990 1 2 3 4 5 6 7 1. Shelters 1.00 2. Hotlines .906 1.00 3. Counseling .906 .995 1.00 4. Children counseling .877 .920 .922 1.00 5. Legal services .646 .723 .736 .661 1.00 6. Batterers counseling .392 .419 .420 .422 .283 1.00 7. Number of referrals .328 .413 .418 .301 .101 -.241 1.00 2000 1 2 3 4 5 6 7 1. Shelters 1.00 2. Hotlines .904 1.00 3. Counseling .900 .981 1.00 4. Children counseling .923 .946 .956 1.00 5. Legal services .851 .921 .959 .933 1.00 6. Batterers counseling .630 .702 .734 .715 .733 1.00 7. Number of referrals -.163 -.144 -.153 -.191 -.246 -.095 1.00

PAGE 119

119 REFERENCES Agresti, Allen. 1996. An Introduction to Categorical Data Analysis. New York: John Wiley and Sons. Allard, Mary Ann, Randy Albelda, Mary Ellen Colten, and Carol Cosenza. 1997. In Harms Way? Domestic Violence, AFDC Receipt, and Welfare Reform in Massachusetts. Boston, MA: University of Massachusetts. Allison, Paul D. 1994. Using panel data to estimate the effects of events. Sociological Methods and Research 23: 179-199. Anderson, Kristin L. 1997. Gender, status, and dome stic violence: An integration of feminist and family violence approaches. Journal of Marriage and the Family 59:655-669. Archer, John. 2000. Sex differences in aggres sion between heterosexual partners: A Metaanalytic review. Psychological Bulletin 126:651-680. Avakeme, Edem F. 1999. Females labor force part icipation and intimate femicide: An empirical assessment of the backlash hypothesis. Violence and Victims 14:277-291. Bailey, William C. 1984. Poverty, inequality, and city homicide rates: Some not so unexpected results. Criminology 22:531-550. Bailey, William C., and Ruth D. Peterson. 1995. Ge nder inequality and violence against women: The case of murder. In Crime and Inequality, eds. John Hagan and Ruth D. Peterson. Stanford, CA: Stanford University Press. Baron, Larry, and Murray A. Straus. 1987. Legi timate violence, violent Attitudes, and rape: A test of the cultural spillover theory. Annals of the New York Academy of Sciences 528:79-110. Berk, Richard A., Sarah F. Berk, Donileen R., Loseke, and David Rauma. 1983. Mutual combat and other family violence. In The dark side of families: Current family violence research, eds. David Finkelhor, Richard J. Gell es, Gerald T. Hotaling, & Murray A. Staus. Beverly Hills, CA: Sage. Barnard, George W, Hernan Vera, Maria I. Ve ra, and Gustave Newman. 1982. Till death do us part: A study of spouse murder. Bulletin of the American Academy of Psychiatry and the Law 10:271-280. Blau, Judith R., and Peter M. Blau. 1982. The cost of inequality: Me tropolitan structure and violent crime. American Sociological Review 47:114-129. Block, Carolyn R., and Anigone Christakos. 1995 Intimate partner homicide in Chicago over 29 Years. Crime & Delinquency 41:496-526.

PAGE 120

120 Block, Carolyn R. 2000. The Chicago Womens Health Risk St udy. Risk of Serious Injury or Death in Intimate Violence: A Collaborative Research Project. Final Report to the National Institute of Justice. Available at http://www.icjia.state.il.us/public/pdf/cwhrs/cwhrs.pdf. Boritch, Helen, and John Hagan. 1990. A century of crim e in Toronto: Gender, class, and patterns of social control, 1859 to 1955. Criminology 28:567. Box, Steven, and Chris Hale. 1983. Liberation a nd female criminality in England and Wales. British Journal of Criminology 23:35-49. Box, Steven, and Chris Hale. 1984. Liberation/Em ancipation, economic marginalization, or less chivalry: The relevance of three theoretical arguments to female crime patterns in England and Wales, 1951-1980. Criminology 22:473-497. Braithwaite, John. 1979. Inequality, Crime, and Public Policy. London: Routledge & Kegan Paul. Browne, Angela. 1985. Assault and homicide at home: When battered women kill. In Advances in Applied Social Psychology, vol. 3, eds. Michael J. Saks and Leonard Saxe. Hillsdale, NJ: Lawrence Erlbaum. Browne Angela. 1986. Assault and homicide at home: When battered women kill. In Advances in Applied Social Psychology, eds. Michael J. Saks and Leonard Saxe. Hillsdale, NJ: Erlbaum. Browne, Angela. 1987. When Battered Women Kill. New York: Free Press. Browne, Angela, and Robert L. Flewelling. 1986. Women as victims or perpetrators of homicide. Presented at the Annual Meet ing of the American Societ y of Criminology, Atlanta, October 29-November 1. Browne, Angela, and Kirk R. W illiams. 1989. Exploring the effect of resource availability and the likelihood of female-p erpetrated homicides. Law and Society Review 23:75-94. Browne, Angela, and Kirk R. Williams. 1993. Gender, intimacy, and lethal violence. Gender & Society 7:78-98. Browne, Angela, Kirk R. Williams, and Donald G. Dutton. 1999. Homicide between intimate partners: A 20-year review. In Homicide: A Sourcebook of Social Research, eds. M. Dwayne Smith and Margaret A. Zahn. Thousand Oaks, CA: Sage. Brush, Lisa D. 1990. Violent acts and injurious outcomes in married couples: Methodological issues in the National Survey of Families and Households. Gender & Society 4:56-67. Byrne, James M., and Robert J. Sampson. 1986. The Social Ecology of Crime. New York: Springer-Verlag.

PAGE 121

121 Bureau of Justice Statistics. 2006. Homicide trends in the United States. Washington, DC. Available at www.ojp.usdoj. gov/bjs/homicide/homtrnd.htm Bureau of Labor Statistics. 2008a. Employment status of the civilian noninstitutional population 16 years and over by sex, 1973 to date. Washington, DC. Available at www.bls.gov/CES Bureau of Labor Statistics. 2008b. Employment status of the civilian noninstitutional population, 1942 to date. Washington, DC. Availabile at www.bls.gov/CES Bursik, Robert J. Jr. 1988. Social disorganiza tion theories of crime and delinquency: Problems and prospects. Criminology 26:519-551. Cameron, Adrian C., and Pravin K. Trivedi. 1998. Regression Analysis of Count Data. Cambridge University Press, Cambridge. Campbell, Jacquelyn C. 1992. If I cant have yo u, no one can: Power an d control in homicide of female partners. In Femicide: The Politics of Women Killing, eds. Jill Radford and Diana E. H. Russell. New York: Twayne Publishers. Cazenave, Noel A., and Margaret A. Zahn. 1992. Women, murder and male dominance: Police reports of domestic violence in Chicago and Philadelphia. In Intimate Violence: Interdisciplinary Perspectives, eds. E. C. Viano. Washington, DC: Hemisphere. Chilton, Roland, and Dee Weber. 2000. Uniform Crime Reporting Program [United States]: Arrests by Age, Sex, and Race for Police Ag encies in Metropolitan Statistical Areas, 1960-1997 [Computer file]. 2nd ICPSR versi on. Amherst, MA: University of Massachusetts [producer]. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor]. Chimbos, Peter D. 1978. Marital Violence: A Study of Interpersonal Homicide. San Francisco: R & E Associates. Cloward, Richard A., and Lloyd Ohlin. 1960. Delinquency and Opportunity: A Theory of Delinquent Gangs. New York: Free Press. Cohen, Albert K. 1955. Delinquent Boys: The Culture of the Gang. New York: Free Press. Crowell, Nancy A., and Ann W. Burgress. 1996. Understanding Violence Against Women. Washington, DC: Research C ouncil National Academy Press. Daly, Martin, and Margo Wilson. 1988. Homicide. New York: Aldine. Daniel, A. E., and P.W. Harris. 1982. Female homicid e offenders referred for pre-trial psychiatric examination: A descriptive study. Bulletin of the American Academy of Psychiatry and Law 10: 261-269.

PAGE 122

122 DeWees Mari A., and Karen F. Parker. 2003. Women, region, and types of homicide. Homicide Studies 7:368-393. Dobash, Russell P., R. Emerson Dobash, Margo Wilson, and Martin Daly. 1992. The myth of sexual symmetry in marital violence. Social Problems 39:401-421. Dugan, Laura, Daniel S. Nagin, and Richard Rosenfeld. 1999. Expl aining the decline in intimate partner homicide: The effects of changing domesticity, womens status, and domestic violence resources. Homicide Studies 3:187-214. Dugan, Laura, Daniel S. Nagin, and Richard Rosenfeld. 2003. Exposure reduction or retaliation? The effects of domestic violence re sources on intimate-partner homicide. Law & Society Review 37:169. Durose, Matthew R., Carolyn Wolf Harlow, Pa trick Langan, Mark Motivans, Ramona R. Rantana, and Erica L. Smith. 2005. Family Violence Statistics. Washington D.C.: Department of Justice, Bureau of Justice Statistics. Ellis, Lee, and Charles Beattie. 1983. The feminist explanation for rape: An empirical test. Journal of Sex Research 19:74-93. Fagan, Jeffrey, and Angela Browne. 1994. Violence between spouses and intimates: Physical aggression between women and men in intimate relationships. In Understanding and Preventing Violence. Volume 3: Social influences, eds. Albert J. Reiss, Jr., & Jeffrey A. Roth. Washington, DC: National Academy Press. Felson, Richard B. 2006. Is violence agains t women about women or about violence? Contexts 5:21-25. Flewelling, Robert L., and Kirk R. Williams 1999. Categorizing homicides: The use of disaggregated data in homicide research. In Homicide: A Sourcebook of Social Research, eds. M. Dwayne Smith and Margaret A. Zahn. Thousand Oaks, CA: Sage. Fox, James Allan. 2005. Uniform Crime Reports [United States]: Supplementary Homicide Reports, 1976-2002 [Computer file]. Compiled by Northeastern University, College of Criminal Justice. ICPSR ed. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [producer and distributor], 2005. Fox, James Allan. 2004. Missing data problems in the SHR. Homicide Studies 8:214-254. Fox, James Alan, and Marianne W. Zawitz. 200 6. Homicide Trends in the United States. Washington, DC: Bureau of Justice Statis tics. Available at www.ojp.usdoj.gov/bjs/ Gallup-Black, Adria. 2005. Twenty years of rural and urban trends in fam ily and intimate partner homicide. Homicide Studies 9:149-173.

PAGE 123

123 Gardner, William, Edward P. Mulvey, & Esther C. Shaw. 1995. Regression analyses of counts and rates: Poisson, overdispersed po isson, and negative binomial models Psychological Bulletin 118:392-404. Gartner, Rosemary, Kathryn Baker, and Fred C. Pampel. 1990. Gender stratification and the gender gap in homicide victimization. Social Problems 37:593-612. Gauthier, DeeAnn K., and William B. Bankston. 1997. Gender equality and the sex ratio of intimate killing. Criminology 35:577-600. Greene, William H. 2000. Econometric Analysis. Macmillin. Goetting, Ann. 1988. Patterns of homicide among women. Journal of Interpersonal Violence 3:3-19. Goetting, Ann. 1995. Homicide in Families and Other Special Populations. New York: Springer. Greenberg, Stephanie W.,William M. Rohe, and Jay R. Williams. 1982. Safe and Secure Neighborhoods: Physical Characteristics and Informal Territorial Control in High and Low Crime Neighborhoods. Washington, DC: National Institute of Justice. Greenburg, David F., and Valeri e West. 2001. State prison populations and their growth, 19711991. Criminology 39: 615-653. Greenfield, Laurence A., Michael R. Rand, Diane Craven, Patsy A. Klaus, Craig A. Perkins, Cherly Ringel, Greg Warchol, Cathy Ma ston, and Fox, James Allan. 1998. Violence by Intimates: Analysis of Data on Crimes by Current or Former Spouses, Boyfriends, and Girlfriends (NCJ-167237). Washington, DC: U.S. Department of Justice. Harris, Anthony R., Stephen H. Thomas, Gene A. Fisher, and David J. Hirsch. 2002. Murder and medicine: The lethality of criminal assault 1960-1999. Homicide Studies 6:128166. Hausman, Jerry, Bronwyn H. Hall, and Zvi Grilic hes. 1984. Econometric models for count data with an application to the patents-R & D relationship. Econometrica 52:909-938. Heimer, Karen. 2000. The nature of crime: Conti nuity and change. Changes in the gender gap in crime and womens economic marginalization. Criminal Justice 1:427-483. Heimer, Karen, Stacy M. Wittrock, and Halim e Unal. 2005. Gender, crime and the economic marginalization of women. In Gender and Crime: Patterns of Victimization and Offending, eds. Karen Heimer and Candace Kruttschitt. New York: New York University Press. Hunnicutt, Gwen, and Lisa M. Broidy. 2004. Liberation and economic marginalization: A reformulation and test of (formerly?) competing models. Journal of Research in Crime and Delinquency 4:130-155.

PAGE 124

124 Johnson, David R. 1995. Alternative methods for the quantitative analysis of panel data in family research: Pooled time-series models. Journal of Marriage and the Family 57:10651077. Junger-Tas, Josine and Ineke Haen Marshall. 1999. The self-report methodology in crime research. In Crime and Justice: A Re view of the Research, ed. Michael Tonry. Chicago: University Press. Jurik, Nancy C., and Russ Winn .1990. Gender a nd homicide: a comparison of men and women who kill. Violence and Victims 5:227-242. Kaukinen, Catherine. 2004. Status compatibil ity, physical violence, and emotional abuse in intimate relationships. Journal of Marriage and Family 66:452. Land, Kenneth C., Patricia L. McCall, and Lawr ence E. Cohen. 1990. Structural covariates of homicide rates: Are ther e any invariances across ti me and social space? American Journal of Sociology 95:922-963. Langan, Patrick A., and Christopher A. Innes. 1986. Preventing Domestic Violence Against Women. Washington, DC: U.S. Department of Justice, Bureau of Justice Statistics. Maddala, G. S. 1983. Limited-dependent and Qual itative Variables in Econometrics. New York: Cambridge University Press. Makepeace, James M. 1989. Dating, livi ng together, and courtship violence. In Violence in Dating Relationships: Emerging Social Issues, eds. Maureen. A. Pirog-Good and Jan E. Stets. New York: Praeger. Makepeace, James M. 1997. Courtship violence as process: A developmental theory. In Violence Between Intimate Partners: Patterns, Causes, and Effect, ed. Albert P. Cardarelli. Boston: Allyn & Bacon. McCall, Patricia L., and Karen F. Parker, K. F. 2005. A dynamic model of racial competition, racial inequality, and interracial violence. Sociological Inquiry 75:273-293. Merton, Robert K. 1968. Social Theory and Social Structure. New York: Free Press. Messner, Scott. F., and Richard Rosenfeld. 1999. Social structure and homicide: Theory and research. In Homicide Studies: A Sourcebook of Social Research, eds. M. Dwayne Smith and Margaret Zahn. Thousand Oaks, CA: Sage. Messner, Steven F., Glenn D. Deane, Luc Anselin, and Benjamin Pearson-Nelson. 2005. Locating the vanguard in rising and fall ing homicide rates across U.S. cities. Criminology 43:661-696. National Coalition Against Domestic Violence. 1991. 1991 National Directory of Domestic Violence Programs: A Guide to Community Shelter, Safe Home and Service Programs. Washington, DC: National Coaliti on Against Domestic Violence.

PAGE 125

125 National Coalition Against Domestic Violence. 1999. 1999-2000 National Directory of Domestic Violence Programs: A Guide to Community Shelter, Safe Home and Service Programs. Denver, CO: National Coalition Against Domestic Violence. Nisbett, Richard E., and Dov Cohen. 1996. Culture of Honor: The Psychology of Violence in the South. Boulder, CO: Westview. Osgood, D. Wayne. 2000. Poissonbased regression analysis of aggregate crime rates. Journal of Quantitative Criminology 16:21-43. Osgood, D. Wayne, & Chambers, Jeff F. 2000. So cial Disorganization outside the metropolis: An analysis of rural youth violence. Criminology 38:81-116. Parker, Karen F. 2004. Industrial shift, polarized labor markets, and ur ban violence: Modeling the dynamics between the economic transf ormation and disaggregated homicide. Criminology 42:619-644. Parker, Karen. F., Patricia L. McCall, and Kenneth C. Land. 1999. Determining social-structural predictors of homicide. In Homicide: A Sourcebook of Social Research, eds. M. Dwayne Smith and Margaret A. Zahn. Thousand Oaks, CA: Sage. Paulsen, Derek J., and Victoria E. Brewer. 2000. The spousal SROK revisited: A comparison of Chicago and Houston intimate partner homicide ratios. Gender Issues 18:88-100. Peterson, Ericka S. L. 1999. Murder as se lf-help: Women and intimate partner homicide. Homicide Studies 3:30-46. Puzone, Carol, A., Linda E. Saltzman, Marcie-Jo M. Kresnow, Matie P. Thompson, and James A. Mercy. 2000. National trends in intimate partner homicide. Violence Against Women 6:409-426. Reckdenwald, Amy, and Karen F. Parker. 2008. The influence of gender inequality and marginalization and types of female offending. Homicide Studies 12:208-226. Reidel Michael, and Joel Best. 1998. Patterns in intimate partner homicide. Homicide Studies 2:305-320. Rodriquez, S. Fernando, and Veronica Henders on. 1995. Intimate homicide: victim-offender relationship in female-perpetrated homicide. Deviant Behavior: An Interdisciplinary Journal 16:45-57. Rosenfeld, Richard. 1997. Changing relationships between men and women: A note on the decline in intimate partner homicide. Homicide Studies 1:72-83. Rosenfeld, Richard. 2000. Pattern s in adult homicide: 1980-1995. In The Crime Drop in America, eds. Alfred Blumstein and Joel Wallman. Cambridge, UK: Cambridge University Press.

PAGE 126

126 Russell, Diane E. 1975. The Politics of Rape: The Victims Perspective. New York: Stein and Day. Saunders, Daniel, and Angela Browne. 2000. Intimate partner homicide. In Case Studies in Family Violence, 2nd, eds. Robert T. Ammerman and Michael Hersen. New York: Plenum Press. Schechter, Susan. 1982. Women and Male Violence. Boston: South End Press. Schwartz, Martin D. 1987. Gender a nd injury in spousal assault. Sociological Forum 20:61-75. Shackelford, Todd K. 2001. Partner-killing by wo men in cohabiting relationships and marital relationships. Homicide Studies 5:253-266. Shackelford, Todd K., and Jenny Mouzos. 2005. Partne r killing by men in cohabiting and marital relationships. Journal of Interpersonal Violence 20:1310-1324. Sherman, Lawrence, and Richard Berk. 1984. Th e specific deterrent effects of arrest for domestic assault. American Sociological Review 49: 261-272. Silverman, Robert, A., and S. K. Mukherjee. 19 87. Intimate homicide: An Analysis of violence social relationships. Behavioral Sciences and the Law 5:37-47. Smith, M. Dwayne, and Victoria R. Brewer. 1992. A sex-specific analysis of correlates of homicide victimization in United States cities. Violence and Victims 4:279-86. Smith, Paige Hall, Kathryn E. Moracco, and John D. Butts. 1998. Partner homicide in context. Homicide Studies 2:400-421. Smith, M. Dwayne, and Margaret A. Zahn. 1998. Homicide: A Sourcebook of Social Research. Thousand Oaks, CA: Sage. STATA. 2003. Cross-sectional time-s eries reference manual. STATA Press. Steffensmeier, Darrell, and Da na Haynie. 2000a. Gender, structural disadvantage, and urban crime: Do macrosocial variables al so explain female offending rates? Criminology 38:403-438. Steffensmeier, Darrell, and Dana Haynie. 2000b. Th e structural sources of urban female violence in the United States. Homicide Studies 4:107-134. Steffensmeier, Darrell, and Cathy Streifel. 1992. Time-series analysis of the female percentage of arrests for property cr imes, 1960-1985: A test of alternative e xplanations. Justice Quarterly 9:77-103.

PAGE 127

127 Stets, Jan E., and Murray A. Straus .1990. Gende r differences in reportin g marital violence and its medical and psychological consequences. In Physical violence in American families: Risk factors and adaptations to violence in 8,145 families. Murray A. Straus and Richard J. Gelles. New Brunswick, N.J.: Transaction. Stout, Karen D., and Patricia Browne. 1995. Le gal and social differences between men and women who kill intimate partners. Affilia 10:194-205. Stark, Evan, Ann Flitcraft, and William Frazier. 1979. Medicine and patriarchal violence: the social construction of a "private" event. International Journal of Health Services 9:461-493. Straus, Murray A. 1993. Physical assau lts by wives: A major social problem. In Current Controversies on Family Violence, eds. Richard J. Gelles and Donileen R. Loseke. Thousand Oaks, CA: Sage. Straus, Murray A., Richard J. Gelle s, and Suzanne K. Steinmetz. 1980. Behind closed doors: Violence in the American family. Doubleday, Garden City, NY. Totman, Jane. 1978. The Murderess: A Psychological Study of Criminal Homicide. San Francisco, CA: R & E Research Associates. U.S. Bureau of the Census. 1994. Census of P opulation and Housing, 1990. Summary Tape File 1a and 3a. Washington D.C.: U.S. Department of Commer ce, Bureau of the Census, 1991. U.S. Bureau of the Census. 2003. Census of P opulation and Housing, 2000. Summary Tape File 1 and 3. Washington, D.C.: U.S. Departme nt of Commerce, Bureau of the Census, 2002. U.S. Department of Justice, Federal Bureau of Investigation. Uniform Crime Reporting Program Data [United States]: Police Employee (LEOKA) Data, 2000 [Computer file]. Compiled by the U.S. Dept. of Justice, Federal Bureau of Investigation. ICPSR ed. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [producer and distributor], 2002. Vieraitis, Lynne M., Marian R. Williams. 2002. Assessing the imp act of gender inequality on female homicide victimization across U.S. cities: A racially disaggregated analysis. Violence Against Women 8:35-63. Villarreal, Andreas. 2004. The Social Ecol ogy of Rural Violence: Land Scarcity, theOrganization of Agricultural Producti on, and the Presence of the State. American Journal of Sociology 110:313-348. Walby, Sylvia. 1986. Patriarchy at Work. Minneapolis, MN: University of Minnesota Press. Whaley, Rachel Bridges, and Steven F. Messner. 2002. Gender equality and gendered Homicides. Homicide Studies 6:188-210.

PAGE 128

128 Williams, Kirk R., and Robert L. Flewelling. 1987. Family, acquaintance, and stranger homicide: Alternative procedures for rate calculations. Criminology 25:543-560. Williams, Kirk R., and Robert L. Flewelling 1988. The Social Production of Criminal Homicide: A Comparative Study of Disaggregat ed Rates in American Cities. American Sociological Review 53:421-431. Williams, Joyce E., and Karen A. Holmes. 1981. The Second Assault: Rape and Public Attitudes. Westport, CT: Greenwood. Wilson Martin. 1989. Spousal conflict and vi olence in evolutionary perspective. In Sociobiology and the Social Sciences, eds. Robert W. Bell and Nancey J. Bell. Texas Tech University Press. Wilson Margo, and Martin Daly. 1992a. Till death do us part. In Femicide: The Politics of Woman Killing, eds. Jill Radford and Diana E. H. Russell. Boston MA: Twayne. Wilson, Margo, and Martin Daly. 1992b. Who kills whom in spouse killings? On the exceptional sex ratio of spousal homicid es in the United States. Criminology 30:189-216. Wilson, Margo, Martin Daly, and Christine Wr ight. 1993. Uxoricide in Canada: Demographic risk patterns. Canadian Journal of Criminology 35:263-291. Wilson, Margo, Holly Johnson, and Martin Daly 1995. Lethal and nonlethal violence against wives. Canadian Journal of Criminology 37:331-361. Wolfgang, Marvin E. 1958. Patterns of Criminal Homicide. Philadelphia: University of Pennsylvania. Yang, Bijou, and David Lester. 1988. The participation of females in the labor force and rates of personal violence (suicide and homicide). Suicide and Life-Threatening Behavior 18:270-78.

PAGE 129

129 BIOGRAPHICAL SKETCH Am y Reckdenwald grew up in Ithaca, New Yo rk. After graduating from high school she moved to Pittsburgh, Pennsylvania where she obta ined her bachelors degree in psychology with a double major in statistics from Carnegie Mellon University. After completing her undergraduate degree in 2001 she moved to Tampa, Florida and later re ceived her masters degree in criminology from the University of South Florida in 2004. In 2008 she completed her Ph.D. in criminology, law and society at the Univer sity of Florida in Gainesville, Florida. After graduating from the University of Florida, Dr. Amy Reckdenwald decided to remain in Florida and accepted a position as an Assistant Professor at Florida Atlantic University, in Boca Raton, Florida.