This month, June 2005, the U.S. Department of Justice, Office of Justice Programs, Bureau of Justice Statistics released their report:

"Family Violence Statistics Including Statistics on Strangers and Acquaintances".

The report was prepared by Matthew R. Durose, Caroline Wolf Harlow,Ph.D., Patrick A. Langan, Ph.D., Mark Motivans, Ph.D., Ramona R.Rantala, and Erica L. Smith, all BJS Statisticians, and they were Assisted by Elizabeth Constantin, BJS Statistical Assistant.

The report is available for free download at
 http://www.ojp.usdoj.gov/bjs.

I hope you download the report and read it because there are problems and the more people who read it the better we will be able to figure out what to do about family violence. I am an admirer of the BJS and believe they are tops in the world in keeping us informed about Justice issues and programs and in providing good data. Yet, I am not
comfortable with the data that are the bases for this report. I believe the report is in error in substance and methods and needs to be corrected quickly.

What concerns me is that the authors of the report know there are serious problems with the data and they still make the statement, that I believe is unjustified, that domestic violence has declined by more than half between 1993 and 2002. I am surprised, quite frankly, that the Women's Groups have not raised a great cry and stink about this report. But, then again, it appears that most, if not all, of the women's groups are not concerned with the reality that is shown by hard data. So, this report reflects the low level of numerical literacy and causal sophistication of the a substantial portion of the expected consumers of this information.

To their credit, the authors do discuss the three fold difference between victim surveys (11%) and police statistics (33%). Please note, differences this large are NOT acceptable for government work. It is unfortunate that the discussion of the data conflicts are not until an appendix (page 71) near the end of the report, and will not be read by most people, or reported by journalists.

We must have closer agreement on robust data before we go into analyzing the minutia of family violence crosstabulations, with variables such as type of weapon, gender, age and familial relationships. Those types of
micro analyses coat our ignorance with a saccharin facade of tables about what is going on with family violence, as well as what might be done about it, tables of data that most likely will not hold up when quality data are obtained. Houston, we have a problem.

Although the report did discuss the reasons why violence was not reported to the police (NB, prepare for your eyes to glaze over during the following paragraphs because too many numbers are reported here without a unifying theme or theory to aid in their assimilation into our weltanschung. There are not even some tasty insights into the nature or causes of family violence to make us feel we honestly have had a
legitimate decrease in family violence and this lack of insights typically indicates that the positive claims are not reliable or substantive):*

"An estimated 17 million violent crimes did not come to the attention of law enforcement officials between 1998 and 2002. These 17 million crimes were 52.8% of all violent victimizations that occurred during those years.

"The two most common reasons for not reporting violence to the police were that the incident was a "private/personal matter" (22.8%) or that the incident was "reported to some other official" (19.5%). An additional 16.4% of victims indicated that the violence was considered "not important enough" to report, while 16.6% did not inform police for
some "other reason."

"Approximately a third of the 1.4 million family violence victims who did not report the incident to police stated the reason for not reporting was that it was a "private/personal matter." A quarter said they did not inform police for some "other reason." Another 12% of nonreporting family violence victims said they did not report the crime in order to "protect the offender."

"Twenty-five percent of the 663,000 victims of spouse violence indicated that they did not notify the authorities because the incident was a "private/ personal matter." "Fear of reprisal" was indicated by 8.3% of nonreporting victims of spouse violence, and an additional 9.3% said they wanted to "protect the offender."

"Thirteen percent of spouse abuse victims indicated they did not report the crime because it was "not important to police." This percentage was higher than the corresponding percentage for violence by a parent against a son or daughter (3%), by other family members (6.3%), by boyfriends or girlfriends (5.5%), by friends or acquaintances (4.2%),
and by strangers (7.7%). In 21.8% of unreported nonfamily violence cases, the incident was considered a "private/personal matter," while in another 20.9% it was "reported to some other official."

"Among nonreporting victims of boyfriend/girlfriend violence, 33.8% said they did not report the incident to police because it was a "private/personal matter." An additional 10.6% indicated "fear of reprisal," and 16.9% stated they wanted to "protect the offender."

"The 7 million victims of violence by a friend or acquaintance who did not report the crime to police most often stated that the incident was "reported to some other official" (29.4%). Twenty-three percent of nonreporting victims of stranger violence indicated that the incident
was "not important enough" to report to police." p. 26.


The analyses of the reasons on failing to report domestic violence are weak -- there were no breakdowns to see how much the failures to report might have contributed to the decrease, by over one half in the violence that the report states happened from 1993 and 2002. It is quite doubtful if
failures to report are distributed randomly. It would have been most helpful here to have had breakdowns by state, county and catchment areas to look at where there were increases and decreases in family violence. The worst consequence of inadequate data collection and analyses is that the good programs, the effective programs that do result in decreases in family violence, are missed.

The data reported here are old and we need to have a system where more current information is collected, analyzed and validated so programs can be initiated and evaluated or simply maintained. One of the major problems with the women's groups is that they have not fought effectively, either in large cities, like Miami or Philadelphia, or
small towns, like St. Augustine and Gainesville, Florida, to have accurate records of family violence easily available to the public for examination, study and discussions. Even with all the personal computers, it is still easier and more emotionally gratifying to picket than to make graphs and tables. There are bright spots that may be of help to us soon as more and more police data is placed on the web, as is being done in Chicago and San Francisco. Unfortunately, it is too soon to see how the data will be used by the public, or the media.

I hope the BJS does a followup on this report and uses more
representative data to discover regression equations that will help show what types of counter - family violence programs are the most effective for different states, cities and regions of the country in truly and reliably decreasing family violence. If nothing else, they (BJS) could set up a system to collect data that are representative and analyze the data to show us what variables, including programs, that might probably be contributing to the decreases, if they exist, and the increases, which some areas are most likely experiencing. With what now appears to be the demise, and it may only be temporary, of the broken windows theory for successful law enforcement (Bernard E. Harcourt & Jens Ludwig: "Broken Windows". Law School, University of Chicago, June, 2005. Download free from SSRN:  http://ssrn.com/abstract_id=743284), it is extremely important that we provide law enforcement with adequate data
and practices on family violence.

The appendix on the problems with the inadequate data, with possible reasons for the data discrepancies, is reproduced below and BJS is to be complemented for publishing this soft caveat to readers. I wish they had been more explicit on the weaknesses of the data. A simple simulation or two to show how having random samples would improve the specificity and the generalizability and the validity of the conclusions
about how much actual family violence is occurring, and has occurred, would have been wonderful. If you read the appendix, and most people will not, please note that it does not reassure you of the reliability or the generalizability or the validity of the data in this report. Small samples and unexplained five fold differences in the amount of measured family violence are not reassuring at all.

The BJS can do better than this report, much better. Of course, it would help if some of the women, maybe even some of the ones who are still attending consciousness raising groups, would let BJS know that these studies are important and volunteer to help establish and maintain data collection in their communites. Such a volunteer program could be effective and if you go to the Betty House, in St. Augustine, Florida, you will see an excellent example of a community working together to provide practical services and basic necessities to victims of family violence, although no one, not even the police, can give you an accurate estimate of how much family violence is occurring here and if the level of violence is increasing or decreasing.

BLS, we are counting on you, and if you look around, you will find support from other places, eg., CDC might be interested in a strategic partnership for determining the true incidence and prevalence of family violence.

Dwight Hines, St. Augustine, Florida



"APPENDIX

"Discrepant findings from two different ways of measuring family violence"

"The extent of family violence in the United States is measured two different ways in this report. One way -- through the National Crime Victimization Survey (NCVS) -- is based on survey interviews with samples of the U.S. population. The other way -- through the FBI's National Incident-Based Reporting System (NIBRS) -- is based on
statistics compiled by police.

"NCVS data summarized in this report indicated that family violence makes up 11% of all violence (both reported to police and unreported), but police statistics summarized in this report indicated that family violence makes up nearly 33% of all police- recorded violence. No firm explanation for this apparent discrep-ancy is offered in the report.
Presumably the discrepancy is caused by a host of factors. The discussion that follows identifies some of them, not necessarily the most important ones.

"One factor pertains to the willingness of victims (and others) to report crime to police. If family violence is more likely than nonfamily violence to be reported to law enforcement authorities, the family violence percentage in police statistics will tend to be larger than the percentage in the statistics for all violence (reported and unreported). To check that possibility, reporting rates for family and nonfamily violence were compared. Results indicated that family violence (59%) is more likely than nonfamily violence (46%) to be reported to police. While the difference is not large enough to fully account for the
11%-versus-33% discrepancy between NCVS and police statistics, results do suggest that a difference in reporting between family and nonfamily violence is a contributing factor.

"A second factor pertains to the fact that the police statistics on family violence used here are not directly comparable to the NCVS statistics in terms of geographical coverage. That is, the police statistics are based on data from agencies reporting in 18 States and the District of Columbia, while the NCVS statistics are for the entire
United States. If the family violence percentage in the 18 States and the District of Columbia happens to be much larger than other States, using these jurisdictions to compare to the NCVS will create the appearance of a greater discrepancy than may actually exist.

"A third factor pertains to the handling of "series" victimizations in the analysis of the NCVS data. The 11% family violence percentage from NCVS data was based on an analysis that treated each "series" victimization (6 or more crimes of a similar nature that the victim is unable to recall individually) as a single victimization. Since family
violence series victimizations comprised 17% (rather than 11%) of all violent series victimizations, treating each series victimization as 6 or more victimizations (rather than 1) would tend to raise the family violence percentage above 11%. To illustrate, when the family violence
percentage was re-computed, this time treating each series using the actual number of incidents (up to 20) indicated by the victim, the percentage rose to 12.5%.

"A fourth factor pertains to the possibility that the NCVS undercounts family violence to a greater extent than nonfamily violence. If so, that would make the family violence percentage look lower than it actually is. The little research on the undercount that has been done -- notably, a small study of assault victims whose victimization was reported to police in San Jose, California, and a national study of victims who went to hospital emergency rooms for their injuries — compared undercounts between family and nonfamily violence. Findings from both studies suggest that the NCVS undercounts family violence to a greater extent than nonfamily violence. The two studies are summarized below.

"San Jose study of known assault victims

"In January 1971, a sample of victims of assault was drawn from the records of the San Jose Police Department. These known victims were interviewed and asked whether they had been a victim of crime at any time in 1970. The victims and their interviewers were not told that the researchers wanted to determine to what extent known assault victims tell interviewers about recent victimizations. Results suggested that family assault victims were less likely than nonfamily assault victims to tell interviewers about their victimization. While 78% of 18 family assault victims did not mention being a victim of family violence, the
comparable percentage for 62 nonfamily assault victims was lower: 29%. P. 71

"The original purpose of the San Jose study was to determine the effects of the passage of time on the recall of criminal victimizations. The survey design emphasized overall reporting issues, not the specific reporting of family violence. Hence, one limitation of the study is the small sample size. Of 126 sampled victims of assault, 18 family assault victims and 62 nonfamily assault victims were located and agreed to participate, a 63.5% response rate.

"Another limitation pertains to the fact that the questionnaire used in the study was more similar to the pre-1993 NCVS questionnaire than the one in use today. Consequently, the San Jose study results may be more
relevant to the pre-1993 version than today’s questionnaire.

"The pre-1993 version was replaced with one that was specifically designed to reduce undercounting of family (and domestic) violence. There is some evidence that the redesign may have succeeded in reducing the undercount. In 1991 (the last full year of interviews using the old questionnaire), family violence victims made up 7.3% of all violence victims (8% if series victimiza- tions totaling 3 are counted as 3
crimes rather than 1; 4 counted as 4; and 5 counted as 5). In 1993 (the first full year using the redesigned questionnaire), the percentage was 10.2%.

"Hospital emergency room study

"While results of the San Jose study only shed light on the extent to which police- reported family violence (family violence that was brought to police attention) is mentioned in NCVS interviews, results of another study may shed light on the extent to which victims tell interviewers about both reported and unreported family violence of a particularly
serious nature. In this other study, two national numbers were compared:

"1. from the NCVS, the total (reported and unreported) estimated number of family violence victims who said they went to a hospital emergency room for treatment for their injuries in 1994

"2. from a national survey — the Study of Injured Victims of Violence (SIVV) — the total (reported and unreported, presumably) estimated number of family violence victims whose injuries brought them to hospital emergency rooms for treatment in 1994.

"All other things being equal, the two numbers should be the same if injured family violence victims tell NCVS interviewers about any incidents that lead to emergency room treatment.

"In fact, SIVV recorded far more emergency room admissions in 1994 than the NCVS, suggesting that the NCVS undercounts victims of serious family violence. The SIVV number of emergency room admissions for family violence (179,000) was found to be over 5 times the number recorded by the NCVS (33,300). By comparison, the SIVV number of emergency room
admissions for nonfamily violence (775,000) was 1.6 times greater than the NCVS-estimate (471,400), suggesting that the NCVS undercounts particularly serious types of family violence to a greater extent than nonfamily violence.

"The relevance of the SIVV study is limited by the fact that its findings pertain to victims who were injured and went to an emergency room. Such victims are a small percentage of all family violence victims. Furthermore, the statistics for the NCVS-documented victims of family violence (who went to emergency rooms) may be unreliable because they are based on a small sample. Also, the SIVV study did not document how many victims had, and how many had not, reported the crime to police. Presumably, the SIVV study included some of both." p. 72.

*Tables omitted.