Allegations of family violence and child abuse in family law children's proceedings

A pre-reform exploratory study
Research Report No. 15 – May 2007

4. Methodology

This study is based on a content analysis of a random sample of 300 court files from the Melbourne, Dandenong and Adelaide registries of the Family Court of Australia (n = 150) and the Federal Magistrates Court (n = 150).80 This chapter describes the detail of the research design.

4.1 Unit of analysis, data collection sites and target year

The unit of analysis in the study was a court file. It is important to note that the progression of each file through the Court can vary from case to case. Most files track the history of a single application and response, or a single set of applications and responses, made at a discrete point in time. Some of these culminate in a final defended hearing, but most are brought to a conclusion before the final hearing takes place. Other files contain successive separate concluded applications filed at different points in time. In children's matters, this might happen where a litigant perceives that there has been a change of circumstances that warrants the Court's reconsideration of parenting arrangements previously ordered.

To analyse files that were as recent as possible but that had been finalised, we selected the 2003 calendar year as the time period of interest. It was hoped that sufficient time had passed for most of these matters to have been finalised by the time the samples had been drawn in May 2006. It is not uncommon, of course, for litigants to return to Court even where "final" orders have been made (Kelly & Fehlberg, 2002).

Registries in South Australia and Victoria were chosen for analysis largely for pragmatic reasons: two barristers who were available for analysing the court files practised in the capital city of each of these states. At the same time, each registry afforded qualitatively different terrains: the Melbourne Registry covers a capital city on the eastern seaboard; by contrast, the Dandenong Registry services Gippsland and the south-eastern regions of Melbourne; and the Adelaide Registry has a state-wide catchment area and, while it encompasses a range of social groups and mixes, over 5% of applications are made by people who identify as being Indigenous.

It is unclear to what extent these registries are representative of other registries in Australia. This caveat is an important one and should be borne in mind when interpreting the study's findings. Courts and their registries vary in culture, demographic environs and resources, and these factors may well be related to the way that allegations and denials of family violence arise and are handled.

4.2 Data collection and coding

Previous studies (such as McIntosh & Prinz, 1993) have often employed undergraduate students to act as coders. In our view, the highly specialised and potentially contentious nature of the present study required a high level of analytic skill, legal expertise, experience of and familiarity with the conduct of court files, and objective independence. For this reason, two barristers were sub-contracted by the Australian Institute of Family Studies to conduct the coding of court files. One barrister was based in Melbourne and had practised in family law for 10 years (8 as a solicitor); the other barrister was based in Adelaide, and had practised civil, commercial, administrative and criminal law for 9 years (8 as a solicitor).

4.3 Sample selection

Two (largely independent) samples were drawn: one from the general population of cases involving children's issues in which an application for final orders had been made (n = 240, hereafter referred to as the general litigants sample); the other comprising a sample of files involving a judicial determination of children's issues (n = 60, hereafter referred to as the judicial determination sample).81 One case file from the judicial determination sample turned out to be out of scope and was subsequently excluded. The final sample thus comprised 59 case files.

The sample of judicial determinations was sought to address the likelihood that a random sample of general population files alone would yield too few judicial determinations for reliable statistical analysis.82 It seemed important to be able to explore allegations and denials of family violence in a sample of judicial determinations because it was unclear whether such allegations and denials would be more prevalent in this sample than in the general population of applications for final orders settled by consent. The Family Court of Australia (2003) noted that:

cases which require a judicial determination may involve complex legal issues, disputes about facts, or serious allegations relating to child abuse or domestic violence. They may also be characterised by unresolved emotional issues which block or delay settlement and require the involvement of an external decision maker. (p. 5)

The general litigants sample was drawn from files that:

  • had commenced in the Family Court or the Federal Magistrates Court in 2003; and
  • concerned children's issues (although not necessarily exclusively children's issues); and
  • had not been the subject of prior litigation about children's issues between the parties; and
  • had since been concluded (that is to say, the applications for final orders had been determined or otherwise dealt with).

The judicial determination sample was drawn using the same selection criteria as the general litigants sample, but with an additional conjunctive condition: that these files were also the subject of a judicial determination of children's issues (rather than a resolution by consent).83

Three decisions were made prior to extracting the final samples. One sampling decision involved opting for an equal number of cases from both the Family Court of Australia and the Federal Magistrates Court.84

The second sampling decision involved stratifying the sample on the basis of geographical location so that the proportion of cases drawn from each registry - Adelaide, Melbourne and Dandenong - would approximate the case load of the Family Court of Australia. This approach sought to improve the representativeness of the sample.85

The third sampling decision involved requesting twice the number of general litigant cases needed (that is, double-sampling so that 600 randomly selected files would be extracted rather than 300).86 In the case of the judicial determination sample, a triple sample was requested.87 This was because judicial determinations can involve: (a) those in which there are no written reasons on file;88 (b) those that contain ex-tempore judgments (judgments made from the bench that may or may not have written reasons on file); and (c) those that contain "reserved" judgments (in which the judge or federal magistrate required additional time to consider the issues and prepare written reasons). The triple sample allowed the research team to estimate the distribution of each of these judgment types and make a decision about the best way to proceed with the sampling. It was decided to ignore the issue of whether there were written reasons available and simply draw a random sample of judicially determined files. This enabled us to estimate the prevalence of allegations and denials of family violence within this small but important group of cases. It also had the benefit of not adding additional layers of complexity to an already complex design.

4.4 Sampling procedures

On the basis of the sampling selection criteria described above, the Manager of the Statistical Services Unit, using the courts' uniform case tracking system Casetrack,89 extracted the target random samples90 and then sent two separate lists of randomly selected target cases to each registry: a primary list and a supplementary list.91 Each list contained roughly twice the number of target cases sought for analysis and included two file numbers for each case: a randomly generated study number and the actual court case file number.92 To protect the privacy of individuals' information, only the randomly allocated study numbers were ever recorded as the unique identifier for each case.

After examining the lists in each registry, the coders reported the number of cases in each primary list to a member of the research team, who in turn arranged for a matching list of random numbers to be generated. Coders used this list to select the relevant number of randomly selected cases from each primary list.93 Where both coders were coding cases in the same registry, as was the case for Melbourne, cases were randomly assigned to coders.94

Each coder proceeded to analyse the selected cases from the main list using a coding frame specifically designed for this study (see below). Coders kept detailed notes on which files were available for analysis - some files were deemed "out of scope"; others were not physically available for a variety of reasons.95 Any shortfall of allocated cases from the main list was then noted, and the required additional cases were sought from the corresponding supplementary list using a similar randomisation process to that used for the main list. For example, if two cases in the main list were found to be "out of scope", four cases (double sampling) were randomly selected from the supplementary list - taking into account the actual number of cases on that particular list (say 4/51 cases). These four cases were then randomly sorted so that the coder could work through the list of target supplementary cases in order.96 Table 4.1 sets out the number of cases sampled from each registry.

Table 4.1 Sampling strategy: Number of cases randomly sampled from each registry by coder
Coder General litigant cases
= 240 (80% of total cases sampled)
Judicially determined cases
= 60 (20% of total cases sampled)
n = 80
(30% of GL)
n = 120
(50% of GL)
= 40
(20% of GL)
n = 20
(30% of JD)
n = 30
(50% of JD)
= 10
(20% of JD)
= 40
n = 40
n = 60
n = 60
n = 20
n = 20
n = 10
n = 10
n = 15
n = 15
n = 5
n = 5
Melbourne coder
n = 150
0 0 40 40 20 20 0 0 10 10 5 5
Adelaide coder
n = 150
40 40 20 20 0 0 10 10 5 5 0 0

Table 4.2 summarises some of the key characteristics of the cases sampled for the four groups of interest: those who went to the Family Court of Australia or the Federal Magistrates Court, with or without a contested hearing.

Table 4.2 Key characteristics of the cases sampled
  General litigants Judgments
n = 120
n = 120
n = 30
n = 29
Case type
% cases involving a couple dispute a 91 97 93 93
% cases involving grandparent/other b 12 6 7 7
Initial applicant or respondent c
% applicant father and respondent mother 33 48 40 38
% applicant mother and respondent father 58 48 53 55
% other 9 4 7 7
Total 100 100 100 100
Couple cases only
Female age (years): mean (standard deviation) 33.8 (6.8) 33.4 (7.6) 34.0 (7.0) 32.5 (8.5)
Male age (years): mean (standard deviation) 37.2 (8.0) 36.3 (8.9) 38.4 (7.5) 35.2 (9.9)
Age symmetry
% where male older than female 74 71 77 68
% where female older than male 13 19 19 28
% where same age 13 10 4 4
Total 100 100 100 100
Relationship status
% married 57 46 61 44
% de facto 35 43 28 49
% unknown 8 11 11 7
Total 100 100 100 100
Duration (median) d
Married: marriage to separation (yrs) 11 7 - -
De facto: separation to filing (months) 6 11 - -
Married: separation to filing (months) 6 10 - -
Children (all cases)
No. of children
% with 1 child 52 50 56 45
% with 2 children 31 35 27 38
% with 3+ children 17 15 17 17
Total 100 100 100 100
Age of youngest child (median) 5 4 3 5
Children's gender e        
% girls only 36 39 24 31
% boys only 36 33 52 45
Both 28 28 24 24
Total 100 100 100 100
Residence of child at initiation of case
% mother 73 75 78 93
% father 9 8 10 0
% other 8 3 3 7
% disputed 3 1 3 0
% split 3 8 3 0
% shared 3 5 3 0
% unknown 1 0 0 0
Total 100 100 100 100
Legal representation (all cases)
At commencement of proceedings
% applicant and respondent 70 57 80 63
% applicant only 19 30 13 22
% respondent only 3 3 7 8
% neither 8 10 0 7
Total 100 100 100 100
At end of proceedings
% applicant and respondent 65 53 53 38
% applicant only 24 26 31 31
% respondent only 2 4 13 10
% neither 9 17 3 21
Total 100 100 100 100
At both start and end of proceedings
% applicant and respondent 59 44 50 37
% neither 6 9 0 7
% other 35 46 50 55
Total 100 100 100 100
Location of parties at initial application
% metropolitan 66 66 57 59
% rural/remote 20 28 40 35
% overseas 1 3 0 3
% not stated 13 3 3 3
Total 100 100 100 100
% metropolitan 54 52 43 38
% rural/remote 19 24 33 42
% overseas 3 4 14 3
% not stated 24 20 10 17
Total 100 100 100 100

Note. aDefined as case where first applicant is former partner of first respondent. bIncludes cases where grandparent or other was second applicant or second respondent. cIn almost all cases, there is only one applicant and one respondent throughout the entire proceeding. dFor de facto couples, there were too few cases with sufficient information (both months and years) to calculate the duration of the relationship; pre-marital cohabitation was often indeterminate because of missing information. No data are presented for the judgment sample because missing information rendered the sample sizes too small. eFor one case, the sex of Child 2 was not identified on court documentation.

Table 4.2 shows that over 90% of cases were disputes between former couples, as would be expected given the nature of disputes on which the sample was based.

Between 48% and 58% of mothers and 33% and 48% of fathers were the initial applicants, while their spouses were the respondents. In three of the four registries, mothers were more likely than fathers to be the applicants in those disputes between parents. Among these former couples, women were, on average aged in their early thirties, while men were in their mid- to late 30s. In most cases, men were older than their former wives.

While close to 60% of parents who filed applications for children's matters in the FCoA had been married to each other, the proportion of parents in the FMC sample who had been married or in de facto relationships was similar (general litigants sample: 46% vs 43%; judicial determination sample: 44% vs 49%). (The marital status of 8-11% of former couples was indeterminate across registries.)

Owing to sample size and missing information, duration data are only available for the general litigants sample. For this group, the median duration between separation and filing was 6 months for those in the Family Court and 10-11 months for those in the Federal Magistrates Court. For married couples, the median length of the marriage (to separation) was 10.5 years for those attending the Family Court, and just under 7 years for those attending the Federal Magistrates Court.

Most of the families in the sample had one or two children, with the average age of the children being 3-5 years old. It is interesting to note that cases requiring a judicial determination - especially those in the FCoA - were more likely than cases in the general litigants sample to involve families in which there were only sons (FCoA: 52%, FMC: 45%, compared with 33-36% in the general litigants sub-samples). At the time the initial application was made, most of the children were living with their mothers (73-93%).

All except a maximum of 10% involved legal representation of one or both parties - most commonly both. Representation of each party was more likely in the Family Court of Australia than in the Federal Magistrates Court (70-80% vs 57-63% involved legal representation of applicant and respondent).

Most of the applicants lived in metropolitan areas. While there was no consistent pattern of residential location for respondents across the four groups, it should be noted that the residential location of between 10% and 24% of cases could not be determined.

4.5 Coding definitions

For the purpose of coding, we adopted the definition of "family violence" contained in s60D (1) of the Family Law Act 1975 (Cth) which, until 30 June 2006, was:

Conduct, whether actual or threatened, by a person towards, or towards the property of, a member of the person's family, that causes that or any other member of the person's family to fear for, or to be apprehensive about, his or her personal well being or safety.

Likewise, for the purpose of coding, we adopted the legislative definition of "child abuse" as:

  1. an assault, including a sexual assault, of the child which is an offence under a law, written or unwritten, in force in the State or Territory in which the act constituting the assault occurs; or
  2. a person involving the child in a sexual activity with that person or another person in which the child is used, directly or indirectly, as a sexual object by the first-mentioned person or the other person, and where there is unequal power in the relationship between the child and the first-mentioned person" (Family Law Act 1975).

Allegations of family violence or child abuse prompted the coding of a flag variable to identify the presence of such allegations in a file.

Drawing on prior work (for example, ABS, 1996; Sheehan & Smyth, 2000), the following types of alleged family violence were recorded:

  1. Sexual assault
  2. Threats of sexual assault
  3. Physical assault: pushed/grabbed/shoved
  4. Physical assault: slapped
  5. Physical assault: kicked, bit or punched
  6. Physical assault: threw anything that could hurt
  7. Physical assault: beaten
  8. Physical assault: hit with something
  9. Physical assault: choked
  10. Physical assault: stabbed/shot
  11. Physical assault: other physical assault
  12. Physical threats of/attempt to: hit with a fist or anything else
  13. Physical threats of/attempt to: stab with a knife
  14. Physical threats of/attempt to: shoot with a gun
  15. Physical threats of/attempt to: other
  16. Verbal abuse
  17. Other emotional abuse
  18. Self-harm
  19. Threatened self-harm
  20. Unspecified abuse
  21. Property damage
  22. Threats of property damage

Correlates of alleged family violence and abuse, as identified in prior work (see previous chapters), were also noted:

  1. Child neglect
  2. Child abduction
  3. Threats of child abduction
  4. Mental illness of other party
  5. Substance abuse (alcohol)
  6. Substance abuse (illicit)

Up to 15 allegations were recorded. Where there were more than 15 allegations, only the 15 most serious allegations were noted. The month and year of each alleged event was noted where it was specified; so too was: which party was the perpetrator; who was the target of the violence or abuse directed at; whether any children witnessed the alleged violence or abuse; how frequent the alleged violence or abuse was ("one-off"; "low": 1-4 instances; "medium": 5-9 instances; "high": 10 or more instances); how long it was alleged to have lasted; how much detail was given; and which document(s) contained the allegation(s).

The following alleged outcomes of the violence or abuse were also noted:

  1. Broken bones
  2. Cuts requiring stitches
  3. Cuts
  4. Scratches
  5. Bruises
  6. Concussion
  7. Miscarriage
  8. Damage to unborn child
  9. Blood nose
  10. Black eye
  11. Other specified physical injury (specified)
  12. Not specified
  13. Restriction(s) on liberty (specified)
  14. Restriction(s) on liberty (unspecified)
  15. Apprehension/fear for self (specified)
  16. Apprehension/fear for self (unspecified)
  17. Fear of property damage (specified)
  18. Fear of property damage (unspecified)
  19. Emotional disturbance/upset
  20. Estrangement from perpetrator
  21. Property damage (specified)
  22. Property damage (unspecified)

The response to each allegation of violence and abuse was coded as:

  1. Complete denial
  2. Confess and avoid (e.g., I did it, but it was mutual/provoked)
  3. "Not admit"
  4. Evidence is silent (does not address it)
  5. Partial denial (e.g., exaggeration)
  6. Deny and counter-allegation of family violence/child abuse
  7. Admit and counter-allegation of family violence/child abuse
  8. Full admission of allegation
  9. Request for further particulars of allegation

It is worth noting that the study did not track the corroboration of denials. First, it is difficult to corroborate most denials. Second, allegations are rarely made when they can be easily disproved (for example, "I wasn't there. I was with ..."). To clarify things, the level of detail in denials was examined. The level of detail given in responding to the allegation was noted with respect to options (1), (2), (5) and (6). They were coded as: high (indicated by a "comprehensive response addressing the particulars of the allegation in detail"); medium (indicated by a "less comprehensive response but providing some detail to rebut the particulars of the allegation"); and low (indicated by a "blanket denial without detail"). The level of corroboration perceived by the Family Report (where conducted), and the source(s) of corroboration cited in the report(s) were also recorded for each allegation.

4.6 Coding frame and computerised data collection tool

A coding frame was developed for the project by the research team, with significant input from the Melbourne barrister, who had extensive experience in family law practice. Where possible, the coding frame required a quantitative rather than qualitative approach in order to minimise the scope for subjectivity in the conduct of the data collection.

The coding frame was first developed as a paper-pencil instrument. However, it soon became clear that the potential complexity of family law proceedings in relation to allegations and denials of family violence would require a more sophisticated approach. As a consequence, a computerised data collection tool was developed using the relational database FileMaker Pro 8. The final coding frame comprised over 4,000 potential data fields.

Upon completion, the coding frame was loaded onto two laptop computers so that each coder could key the relevant data from case files directly into the customised software. A sample of 25 cases was used to pilot test the coding frame. None of the cases in this sample was part of the final selected sample. Minor refinements were made to the software on a rolling basis until the coding frame was able to deal with all of the relevant information contained in the pilot sample.

There were many advantages in using a computer-assisted coding frame: (a) it was easy to enter data in a systematic way into the system, without unnecessary typing; (b) with allegations and denials of family violence forming the unit of analysis in the data collection process, information related to each allegation (such as corroboration) could be tracked horizontally in tables across different sections; (c) the computer software allowed data logic and validity checks to be included which, in turn, improved the integrity of the data; (d) data security was enhanced because all of the information was stored in two password-protected laptops, both of which were almost always kept within the Court for the duration of the data collection phase; and (e) the data could readily be exported directly into a statistical analysis package (keying data from a paper-pencil version into a statistical package is likely to lead to keying and transposition errors).

Figures 4.1 through 4.6 contain examples of the data capture screens.

Figure 4.1 Coding frame: Menu screen

Figure 4.1 Coding frame menu screen, described in text.

Figure 4.2 Coding frame: Flag question screen

Figure 4.2 Coding frame Flag question screen, described in text.

Figure 4.3 Coding frame: Allegations by Applicant screen

Figure 4.3 Coding frame Allegations by Applicant screen, described in text.

Figure 4.4 Timing of allegations: Sample screen

Figure 4.4 Timing of allegations Sample screen, described in text.

Figure 4.5 Form of denial (if any) of allegation: Sample screen

Figure 4.5 Form of denial of allegation Sample screen, described in text.

Figure 4.6 Coding frame: Corroboration screen

Figure 4.6 Coding frame Corroboration screen, described in text.

Table 4.3 summarises the content of each section of the coding frame. This summary is necessarily an over-simplification of each section.

Table 4.3 Content of coding frame
  Section Content
1 Case info Coder's name, case ID, registry and date examined.
2 The parties Parties' ages, occupations and relationship chronology.
3 Children Details about children: ages, parentage and usual residential care.
4 Proceedings Details of what orders were sought by each party in each set of proceedings (interim and final).
5 Proceedings representation Were the parties legally represented at the commencement and the conclusion of proceedings? Was a child representative appointed?
6 Flag question Was an allegation of family violence or child abuse raised at any point?
7 Other proceedings Were any other proceedings noted (such as family violence, child protection and child support etc.)?
8 Nature of allegations - Applicant What type of violence or abuse did the applicant allege? When did it allegedly occur? Who was the alleged perpetrator? Who was the conduct directed at? Was it seen/heard by any of the children? If so, which ones? What was the frequency and duration of the alleged violence/abuse? How detailed was each allegation?
9 Nature of allegations - respondent As above (8), but addressed to the allegations of the respondent.
10 Applicant's alleged outcomes of the violence Applicant's alleged outcomes of the violence or abuse perpetrated by the respondent (e.g., bruises, cuts, broken bones, fear, apprehension)
11 Respondent's alleged outcomes of the violence Respondent's alleged outcomes of the violence or abuse perpetrated by the applicant (e.g., bruises, cuts, broken bones, fear, apprehension)
12 Timing of applicant's allegations For each allegation: When was the allegation of violence and/or abuse first raised by the applicant? Reasons given, if any, for a delay in reporting? If made prior to proceedings, to whom?
13 Timing of respondent's allegations As above (12), but addressed to the allegations of the respondent.
14 Applicant's evidence of corroboration Was there any corroborating evidence? If so, from whom and what form did it take?
15 Respondent's evidence of corroboration As above (14), but addressed to the allegations of the respondent.
16 Subpoenas Were any subpoenas on file? When was it filed, by whom, and to whom was it addressed?
17 Form 4 Was a Form 4 filed? If so, by whom? Who was the alleged perpetrator? Who was the alleged victim?
18 Child protection agency involvement Did the court file indicate the involvement of a state child protection agency in the family law proceedings? If so, what did the agency do? What concerns, if any, did the agency express to the court?
19 Respondent's response to allegations Did the respondent respond to any of the allegations and, if so, how (e.g., full admission, full denial, partial denial or was the evidence silent)?
20 Applicant's response to allegations As above (19), but addressed to applicant's response to allegations.
21 Family Reports Was there a Family Report for an interim hearing? What was the origin of the report?
22 Corroboration in interim report of applicant's allegations What level of corroboration was contained in the report with respect to allegations by the applicant (e.g., fully corroborated, partially corroborated, fully discredited, partially discredited, no opinion offered or not raised)? What was the source of corroboration cited? Was there a link between the corroboration and reporter's final recommendations?
23 Corroboration in interim report of respondent's allegations As above (22), but addressed to respondent's allegations.
24 Reporter's unconditional recommendations - Interim What were the reporter's unconditional recommendations (interim report)?
25 Reporter's conditional recommendations (true) - Interim If the report writer's recommendations were conditional on court finding that the allegations were true, what were those recommendations?
26 Reporter's conditional recommendations (false) - Interim If the report writer's recommendations were conditional on court finding that the allegations were false, what were those recommendations?
27 Final hearing Was there a Family Report for a final hearing?
28 Corroboration - Final report: Applicant's allegations What level of corroboration (if any) was contained in the final report with respect to allegations by the applicant?
29 Corroboration - Final report: Respondent's allegations What level of corroboration (if any) was contained in the final report with respect to allegations by the respondent?
30 Reporter's unconditional recommendations - Final What were the reporter's unconditional recommendations (final report)?
31 Reporter's conditional recommendations (true) - Final If the report writer's recommendations were conditional on court finding that the allegations were true, what were those recommendations?
32 Reporter's unconditional recommendations (false) - Final If the report writer's recommendations were conditional on court finding that the allegations were false, what were those recommendations?
33 Court orders What were the substantive orders of the court (interim and final)?
34 Evidence cited to support applicant's allegations Where the court made findings about violence and abuse alleged by the applicant, what evidence did it cite in support of those findings?
35 Evidence cited to support respondent's allegations As above (34), with respect to the respondents' allegations.
36 Court findings Where the allegations were the subject of a finding, to what extent did the judge/federal magistrate expressly link the finding about violence or child abuse with the outcome ordered? Did the judge/federal magistrate make a finding that there was an "unacceptable risk" to the welfare of a child/children? What was the nature of the risk identified by the judge/federal magistrate (e.g. sexual abuse, exposure to violence, etc.)? What actions or orders were made or taken against a person found to have made a false allegation or false denial?

4.7 Training and inter-rater agreement

A 2-day training workshop was held early in the project in which all members of the team met to clarify the project's aims and parameters. The two coders simultaneously analysed the first five files in the Melbourne Registry of the FCoA in order to gain consistency in the way that they applied the various codes. They then coded separate files but used each other as a "sounding board" to deal with queries and ambiguities in the data. One of the members of the research team acted as an impartial third person in this process.97

Ten cases were coded by both coders in the early stages of the content analysis. This sample was used to check the consistency of each coder's coding. Some anomalies surfaced and were corrected by three-way discussion.98

4.8 Timelines and logistics

The project was conducted under tight timelines, given the size and complexity of the study. Eight weeks were allocated to obtain all of the necessary approvals99; 12 weeks were allocated for the content analysis of the 300 court files (which included two separate extractions of the various court samples); and 12 weeks were initially allocated for analysis and final write-up.

Once the coding frame was finalised and the random cases selected the coders began work on the selected files. There was considerable variation in the length of each file - some files comprised just a few pages - while at the other end of the spectrum, one file required five large boxes to hold its material.

4.9 Caveats and limitations

A number of caveats warrant mention. First, the samples were small - particularly the judicial determination sub-samples - which increases the likelihood of error in our estimates. Moreover, since the registry locations were selected on the basis of purposive sampling, the findings may not be generalisable to other registries in Australia.

Second, no information was sought from state- or territory-based child protection agencies, the police or local courts exercising power under the Family Law Act 1975. As noted earlier, state jurisdictions act as the frontline for dealing with allegations of family violence and child abuse in Australia. While very basic information was noted in the present study about the involvement of such agencies in each case, the omission of state-based data (and the potential interrelationship with the Family Court and Federal Magistrates Court data) means that the scope of the study was far from comprehensive. That said, the inclusion of such data represents an ambitious project and is well beyond the scope of the present investigation.

Third, it is important to note that the findings are based entirely on the information documented in the court file. While the courts expect litigants to provide most of the legal "evidence" in the form of written affidavits that are to be found on the court file, the file remains limited as a source of information about what has transpired in a relationship. For example, the court file does not retain documents produced to the court in answer to a subpoena; yet the subpoena process is frequently used in litigation to corroborate allegations of violence. No interviews were conducted with any of the parties involved in the disputes, or with any third parties (including legal representatives) who may have had other information of relevance. The study focused on the prevalence of allegations in selected registries and not the prevalence of violence per se.

Fourth, some allegations may be an underestimate of the real situation, some may be an overestimate, and some (whether made deliberately or otherwise) may be untrue. This study did not attempt to assess the veracity of allegations on the basis of court records, especially the extent to which any of these statements had been deliberately fabricated. Assessing veracity would require a much more sophisticated research design involving a multi-disciplinary approach within multiple agencies within state/territory and federal governments.

Fifth, it is also important to note that allegations (or non-disclosures) are typically filtered through the lens of litigants' legal advice and/or representation. Legal advice and/or representation therefore act as a protective screen enveloping the data.

Finally, there are likely to be additional considerations for certain groups of individuals (such as people from non-English speaking backgrounds, Indigenous people, people with disabilities, and people in same-sex relationships). This study is not well placed to identify these considerations because of the relatively small number of people from such groups included in the samples. More focused qualitative approaches are likely to be better suited to exploring the considerations for these particular groups.

All of these caveats mean that great care should be taken when interpreting or reporting on these data. The study's sophisticated research design should nonetheless provide useful pre-reform data against which the recent family law reform package can be assessed at a later date.

It is worth noting that while a range of sampling strategies have been used in prior research to estimate the prevalence of allegations of family violence or allegations of child abuse, no study has appeared to draw random samples from two naturally occurring populations of family law court cases - namely, the general population of applications for children's matters, and cases requiring a judicial determination - in order to estimate the prevalence of family violence and child abuse and their co-occurrence in each sample.100 This sampling strategy is the backbone of this study and one of its unique features.

We now turn to the study's findings.

  • Chapter 5 describes the frequency, nature and source of allegations of family violence and child abuse, and the extent to which children were reported as having witnessed violence.
  • Chapter 6 sets out the extent to which allegations were corroborated by the alleging party, and were denied, admitted or left unanswered by the other party.
  • Chapter 7 presents data on the relationship between allegations and court outcomes with respect to post-separation parenting arrangements.


80 The target sample of 300 cases was chosen on the basis of statistical power, efficiency and cost in terms of (data collection) time and money.

81 One case overlapped both samples. This case was excluded from the judicial determination sample, and replaced by another randomly selected case from that sample. The general population sample was analysed before the judicial determination sample.

82 We were expecting around 5-6% of cases in which final orders were sought to involve a judicial determination. A random sample of 300 cases would thus have been likely to have produced only 15-18 cases for analysis, which would have had little statistical power.

83 The judicial determination sample was drawn about 6 weeks after the general population sample was drawn.

84 This decision was made for several reasons. First, very little is known about family law matters in the FMC. Second, in discussions with the courts and other stakeholders, the view was expressed that there would be a number of cases in the FMC in which there were allegations of violence. Third, increasing the number of cases from the FMC would increase the statistical reliability for the FMC data without substantially reducing it for the FCoA. Fourth, there is no defensible basis for not having an equal number of cases from each court. Initially, we were intending to sample 2/3 of the 300 files from the FCoA, and 1/3 from the FMC, but decided to opt for an equal split for the reasons just outlined.

85 We chose this regional distribution based on the Family Court of Australia's records of applications filed in the Adelaide, Melbourne and Dandenong registries in 2002/2003 (as summarised on the FCoA's website).

86 We did this for three reasons: (a) it makes the sampling much simpler and reduces the risk of problems associated with a non-random sample; (b) it is more efficient from the Court's perspective; and (c) it meant that there would be no delays in obtaining an additional sample (or cases) should it be required. This decision turned out to be prudent because some cases were deemed to be "out of scope" or were not available for analysis.

87 In some cases, a smaller master sample was received because there were simply not enough judicial determinations under the specified conditions at a registry in the time period of interest.

88 This may have been because the judge or federal magistrate delivered reasons for their decision orally in Court and it was not considered necessary in all of the circumstances to have those reasons transcribed into a written judgment.

89 Casetrack is the common administrative data tool used by both the Family Court of Australia and the Federal Magistrates Court to track court files.

90 The methodology used by the Court for selecting the final sample was as follows. The population consisted of all finalised final orders applications lodged with the FCoA or FMC during 2003 where the file containing the application was, at close of business 3 May 2006, in either the Adelaide, Dandenong or Melbourne FCoA or FMC registries, and the applicant indicated "children's issues" or "children and financial issues".

Each record was assigned a random number. The random number was generated in SAS using the ranuni function, with a seed chosen at random from the book A million random digits with 100,000 normal deviates (see The population was then sorted in ascending order by the generated random number.

Before selecting the sample, the population was treated as six distinct entities (one for each location and jurisdiction). Each nth record was then selected, where each n was calculated by dividing the number of units in the population for each location and jurisdiction by the size of the sample in that location and jurisdiction and rounding to the nearest integer.

91 The supplementary list was to facilitate the selection of additional cases should this be necessary.

92 These lists acted as the case ID "key" and were kept at each registry so that the coders could always cross-check study numbers with actual case numbers.

93 Bruce Smyth coordinated this process. Ross Millward at the Australian Institute of Family Studies conducted the random allocation of these lists, using Excel to generate a random sample of n cases, which was then sorted in random order. This process was necessary because systematicity was present in the sampling frame: each list was sorted by the date that each application for final orders had been originally filed; without random allocation across the sampling frame, cases from a particular period may have been over-sampled (e.g., the first half of the list would have led to analysing cases from only the first half of 2003).

94 Both coders analysed cases in the Melbourne Registry of the FCoA and FMC (see Table 4.1). For logistical reasons, the Adelaide-based coder analysed cases in the Adelaide Registry and the Melbourne-based coder analysed cases in the Dandenong Registry.

95 "Out of scope" files included cases that: revealed prior litigation between parties concerning children's issues, remained "on foot" (that is, were still active), or had been transferred to or returned to another court.

96 This was necessary because files on the supplementary list could also be deemed "out of scope" or not available.

97 Lawrie Moloney fulfilled this role. He was a former Director of the Family Court Counselling Service in Victoria during the first 10 years of its existence.

98 These anomalies could not be quantified because of the complexity of the data structure; the number of actual data points varies for each case, and cannot be readily averaged into a denominator.

99 Approvals needed to be obtained from the Institute's Ethics Committee; the Family Court's Research and Ethics Committee; the Federal Magistrates Court; and from the Attorney-General. In addition, several discussions were also held with the Privacy Commission in order to check that a Public Interest determination was not required - it was not.

100 As noted in Chapter 2, Sorenson et al. (1995) sampled 60 "contested custody" cases from seven judicial circuits throughout Florida; and Smart et al. (2003) drew a random sample of 430 residence and contact disputes from three county courts in England.