Methodology

Population and period of observation
Unit of analysis: the court career
Record selection, matching and weighting
Population at risk
Criminal charges heard in municipal courts in Quebec
Classification of offences and incidents

Population and period of observation

The population of the study consists of all persons born between April 1, 1979 and March 31, 1980¹, who had at least one charge which was related to a federal statute offence allegedly committed before their 22^nd birthday, and was disposed of in youth court or adult provincial (criminal) court between April 1, 1991 and March 31, 2003, in one of the following jurisdictions: Newfoundland and Labrador, Prince Edward Island, Quebec, Ontario, Saskatchewan, and Alberta. The period of observation was from the 12^th birthday to the day before the 22^nd birthday; i.e. charges were selected for which the age of the accused on the date of the alleged offence was between 12 and 21 inclusive.

Prospective studies of criminal careers generally choose a birth cohort — that is, a population who were all born during a given period of time, often a period of one year — and track this cohort’s criminal behaviour over as long a period of time as possible. In the present study, the choice of birth cohort involved a trade-off between the completeness of coverage across Canada, and the number of years over which cohort members could be tracked.

Tracking court careers in Canada requires records from two surveys: the Youth Court Survey (YCS), which covers accused aged between 12 and 17 years at the time of the alleged offence, and the Adult Criminal Court Survey (ACCS), which covers accused aged 18 years or older. The ACCS has covered, since fiscal 1994/95, provincial (criminal) courts, excluding superior courts, in the following provinces and territories: Newfoundland and Labrador, Prince Edward Island, Nova Scotia, Quebec, Ontario, Saskatchewan, Alberta, and the Yukon. The YCS has covered all provinces and territories since 1991/92. Prior to 1991/92, Ontario did not report to the YCS. It was considered essential to include Ontario in the study, because of its size; thus, the earliest possible observation year was 1991/92. Since 12 years old at the time of the alleged offence is the minimum age of jurisdiction of the youth courts, the earliest available YCS data year — fiscal 1991/92 — implies using a cohort born in 1979/80, who would all have their 12^th birthdays during fiscal 1991/92.

At the time when the research was conducted, fiscal 2002/03 was the latest year for which YCS and ACCS data were available. Thus, the selected observation period was from 1991/92 to 2002/03, and the selected population was: all persons born in 1979/80 who had charges heard in youth court or provincial criminal court in Newfoundland and Labrador, Prince Edward Island, Quebec, Ontario, Saskatchewan, and Alberta.² The selection of jurisdictions was determined primarily by the limitations of the coverage of the ACCS, but Nova Scotia was also omitted because records for Nova Scotia for 12 to 15 year olds could not be matched with those for 16 to 17 year olds, and the Yukon was omitted because of record matching problems.³ The six provinces included account for approximately 78% of the population of Canada during this time period.

The period of April 1, 1991 to March 31, 2003 for which charge data were available corresponds to varying periods in the lives of members of the cohort, depending on which day of the year they were born. The oldest members, born on April 1, 1979, turned 12 on April 1, 1991, and were a day less than 24 years old on March 31, 2003. The youngest members, born on March 31, 1980, would have been a day past their 11th birthdays on April 1, 1991, and would have had their 23^rd birthdays on March 31, 2003. They would not have been “at risk” of being charged until their 12^th birthday, since 12 years was the minimum age of criminal responsibility in Canada during the observation period. Thus, members of the cohort would have had different lengths of time at risk if all were observed to March 31, 2003.

Instead, all individuals were tracked to the day before their 22^nd birthday.  The choice of the 22^nd birthday as the cut-off date was based on an analysis of the length of time which it normally took for charges to be processed to disposition.  Charges are reported to the YCS and ACCS when they are disposed of; that is, when their court process is completed. Therefore, YCS and ACCS data do not include all charges related to offences committed during the reporting year — only those charges which are disposed of during the reporting year. There are many reasons why charges related to an incident occurring in a given year may not have their first hearing in court until a subsequent year: for example, there may be a delay in the incident coming to the attention of police (as, for example, with some cases of child sexual abuse), the police investigation may be lengthy, or the charges laid or recommended by police may be screened by the Crown. In addition, court processing of many charges continues into succeeding years. Thus, charges related to offences allegedly committed at the ages of, say, 22 or 23, by some, but not all, members of the cohort born in 1979/80 would not have been disposed of by March 31, 2003, and would not appear in the data for 2001/02 or 2002/03. Therefore, it was decided to limit the incidents included in each person’s career to those allegedly occurring before the 22^nd birthday, to minimize the underestimation of the level of offending in the last year or two of observation.

Table 12 shows the flow through the youth courts and criminal courts of disposed-of charges associated with incidents involving persons born during fiscal 1979/80. Panel (a) gives raw numbers of incidents, panel (b) shows row percentages, and panel (c) shows cumulative row percentages. Each row represents incidents which occurred during a given fiscal year. Each column shows the reporting year in which the charges related to the incident were disposed of, and reported to the YCS or ACCS. For example, the first row of panel (a) shows the number of incidents during 1991/92, when the 1979/80 birth cohort had their 12^th birthdays. The first row of panel (b) shows that charges related to 16% of these incidents were disposed of during 1991/92, 64% during 1992/93, 16% during 1993/94, and the rest during 1994/95 and 1995/96. (In fact, as panel (a) shows, charges from a few incidents which occurred in 1991/92 continued to be reported as disposed of in every year until 2002/03, but the numbers were so small that they rounded to 0% in panel (b)). The first row of panel (c) shows that charges related to 80% of incidents which occurred in 1991/92 were disposed of within one reporting year (1992/93) in addition to 1991/92 itself, and that charges related to 96% of incidents were disposed of within two additional reporting years.

Of course, the validity of these percentages depends on the assumption that charges related to all, or practically all, incidents occurring in a given year were disposed of by 2002/03, in order to have a correct total number of incidents to be used as the denominator in calculating the percentage. The row totals in panel (a), and the patterns of cumulative percentages in panel (c), suggest that this assumption is reasonable for incidents occurring up to 2000/01, but after that, the percentages become less reliable, as the charges related to more and more incidents are presumably still outstanding after 2002/03.

On the basis of the distributions of incidents which occurred up to 2000/01, we can conclude that charges related to 74 to 85% of incidents will be included in the data if one reporting year in addition to the year of occurrence of the incident is used, and that 94 to 98% of incidents will be included if two additional years are used. Consequently, the career was defined as all incidents which occurred before the 22^nd birthday. Since members of the cohort had their 22^nd birthdays during 2001/02, then charges related to something like 15 to 26% of their incidents which occurred at the age of 21, and 2 to 6% of incidents which occurred at the age of 20, would not have been disposed of by the end of the observation period (March 31, 2003). This under-estimate of age-specific offending at the ages of 20 and 21 was considered to be an acceptable trade-off against the value of being able to follow criminal careers during these two years.⁵

Unit of analysis: the court career

All charge records which fit these criteria were extracted from the data files of the Youth Court Survey and the Adult Criminal Court Survey for fiscal 1991/92 to 2002/03. Charges pertaining to the same person were linked to create a record of his or her court career up to the age of 21 inclusive. The analysis of court careers requires the construction of a data file with the person (or, equivalently, the career) as the unit of analysis (data record). YCS and ACCS data files use the charge as the basic data record. Thus, charge records from the YCS and ACCS which pertained to the same person were aggregated to form a person record which contained summary variables capturing the relevant attributes of the constituent charges.

However, in research on criminal careers, the unit of count is not the charge, but the (alleged) “crime”: that is, the act which constitutes a violation of the law, and which may result in more than one charge, or alleged offence. The “crime” is usually operationalized in criminal careers research by the criminal incident, also known as the occurrence, offence, or offence-episode. Perhaps this is because most criminal careers research uses data from police records or self-report surveys, which usually use the incident as the unit of analysis. Thus, in aggregating charge records to the person file, an intermediate level of aggregation — the incident — was used, in order to be able to construct person-based summary variables such as the number of incidents in the person’s court career, which would be of much more substantive interest than the number of charges. The incident was defined operationally as all charges (i.e. alleged offences) pertaining to the same person and having the same date of offence.⁴

Matching of charge records for the same person was not straightforward, since there is no unique person identifier in the YCS or ACCS. Matching must be done using the province, person’s name, date of birth, and sex, and any other relevant fields. This raises the issue of potential false positive matches. Different people have the same name, date of birth and sex. Furthermore, the accused person’s name is not recorded as such in the YCS and ACCS – it is encoded in a 4-character Russell Soundex code (or Henri code in Quebec), which is not unique: many names are encoded with the same Soundex or Henri code.⁶ Thus, matching on the Soundex or Henri code, date of birth and sex could result in false positive matches: records for different people would be erroneously treated as pertaining to a single person. The result would be an underestimate of the number of unique offenders and an overestimate of the numbers of incidents in their careers.

This is not an issue when aggregating charge records into incidents, because that matching process also uses the date of offence: for a false positive to occur, it would be necessary for two different people with the same Soundex code, date of birth, and sex, to be charged in connection with incidents occurring on the same day in the same province. The probability of this is negligible. However, false positives are a more serious issue in the matching of incidents to construct the person record, since that matching process relies on only the province, Soundex or Henri code, date of birth, and sex.

False negatives — where two records should be matched but are not — are also a potential problem in record matching. A false negative could occur if court records contained more than one name for the same person; for example, if a person changed his or her name during the observation period,⁷ or used an alias, or if the name was misspelled. An incorrect record of the date of birth or sex would also result in a false negative. A false negative would also occur if the person committed crimes in more than one province, since all matching was done within provinces. Matching could have been done across the entire set of provinces which constituted the study population, in order to maintain the integrity of careers which crossed provincial boundaries, but this would have introduced two other problems:

• Since court data from Quebec use the Henri code, these records could not have been matched with records for other provinces;
• Matching within several provinces would have greatly increased the size of the “pool” of persons being matched, which would have exacerbated the problem of false positives (see below for the relationship between the size of the pool and the probability of false positives).

An analysis of the probability of false positive matches was conducted by determining the rate of occurrence of each Soundex or Henri code in the populations of the provinces in the study, using electronic telephone directories. This enabled the calculation, for each Soundex and Henri code, the expected rate of false positives, when it was used for matching in combination with birth date and sex. Soundex and Henri codes vary greatly in their vulnerability to false positive matches, since some encode very common names and others do not.

The probability of false positives is directly related to the number of records being matched, which is approximately proportional to the population of the geographical area, and the number of years, within which matching is being done. There would be many false positives if records for many years for all of Canada were being matched, and few or none if records were matched for only a few years within one town. Thus, in a study such as the present one, where the number of years of matching is fixed (1991/92 to 2002/03), the “match quality” or “match efficiency” (i.e. non-vulnerability to false positives) of Soundex and Henri codes is related both to the commonness of the names which they encode, and to the population of the area within which matching is being done.

On the basis of this quality analysis, four categories of the “quality” of Soundex and Henri codes were defined:

• 0 – The code is rare enough that there is 99% or better match efficiency rate.
• 1 – 95% – 99% match efficiency rate.
• 2 – 90% – 95% match efficiency rate.
• 3 – less than 90% match efficiency rate.

“Match efficiency” refers to the absence of false positives; e.g. 99% match efficiency means that 1% of matches are expected to be false positives, and “99% or better” means that 1% or fewer false positives are expected.

Records (and therefore persons) whose Soundex or Henri codes had worse than a 95% match efficiency (i.e. quality codes of 2 or 3) were eliminated from the study. The rationale for simply eliminating these records is that, as a record selection criterion, Soundex and Henri codes (representing persons’ names) are presumed to be unbiased with respect to criminal behaviour. A person with a common name such as John Smith, whose Soundex would probably have a quality code of 3, is no more or less likely to have a criminal career, or a career with particular characteristics, than a person with an uncommon name. Thus, the records with Soundex and Henri quality codes of 0 and 1 constitute a subset which is presumed to be representative of the entire population with respect to the phenomenon under study (criminal behaviour).

Record selection, matching and weighting

If a sample is small relative to the population, then even very small selection biases can result in significant unrepresentativeness. Also, even perfectly random samples with small numbers suffer from serious random sampling error and unstable parameter estimates based on small cell sizes. Table 13 shows the proportions of incident records with Soundex quality codes of 0 or 1 in the YCS and ACCS for 1991/92 to 2002/03, for the 1979/80 birth cohort.

These proportions are all acceptable, except for Ontario and Alberta, where approximately two-thirds and one-half of the records have unacceptable Soundex quality codes. This problem was resolved by “regionalizing” Ontario and Alberta. Just as Canada had already been “regionalized” by matching records within provinces rather than across the whole country, so Ontario and Alberta were subdivided into regions, with record matching done within each region. The effect of this regionalization (like the effect of matching within provinces instead of Canada as a whole) is to reduce the number of records within which matching is being done, and therefore the likelihood of false positives, while increasing the probability of false negatives by making it impossible to match across regional boundaries within a province. The choice of the number of regions within each province is thus based on the trade-off between reducing regional populations sufficiently to produce a high proportion of records with Soundex quality codes of 0 or 1, while not making the regions so small that people are likely to commit crimes in more than one region during the 10-year observation period. Ontario was divided into four regions, and Alberta into two. Regional boundaries were selected in order to make the number of records for each region as close as possible to equal (and therefore all minimal), while at the same time defining regions which were socio-economically as integral as possible, in order to minimize the probability that people would commit crimes across regional boundaries. Ontario was divided into: the Toronto Census Metropolitan Area (CMA), southern Ontario west of the Toronto CMA, southern Ontario east of the Toronto CMA, and northern Ontario. Alberta was divided into northern and southern regions, based on the Edmonton and Calgary CMAs respectively. The results of the regionalization are shown in Table 14. The proportions of incident records to be retained for Ontario and Alberta are now acceptable.

The records with Soundex quality codes of 0 or 1 were then aggregated into person (career) records. To compensate for the deletion of records with Soundexes of 2 and 3, each person record was assigned a weight, which was the inverse of the selection fractions shown in Table 14. For example, 73.5% of incident records for males in Saskatchewan had Soundex codes of 0 or 1. Therefore, the person records for males in Saskatchewan were weighted by the inverse of 0.735, which is 1.36. Individuals whose first referred incident occurred after their 22^nd birthday were eliminated from the population.  The weighted numbers of offenders in each province are shown in Table 15. These are the offender populations on which the analyses are based.

Population at risk

The prevalence of, or participation in, offending or alleged offending is usually expressed in the criminal careers literature as the proportion of the cohort who (allegedly) committed an offence at a given age (age-specific prevalence), or ever (allegedly) committed an offence up to a given age (age-specific cumulative prevalence), or ever (allegedly) committed an offence during the period of observation (overall or lifetime prevalence). Calculation of such prevalence estimates requires both the number of persons who (allegedly) exhibited the behaviour, and the number of persons at risk of exhibiting it — the eligible population at risk.

Using the YCS and ACCS data, one cannot track exactly the same group of individuals for ten years — from their 12^th birthday up to their 22^nd birthday. Each year, some individuals will either immigrate to or emigrate from Canada or the parts of Canada included in the study, and/or will move between the provinces and territories under study. Consequently, determining the exact total eligible population at risk of a contact with the court system is not possible. However, population data provided by Statistics Canada for each age and sex in Canadian provinces and territories may be used to approximate the male and female population born between April 1st, 1979 and March 1st, 1980 for each year while under the jurisdiction of the Canadian justice system.

The total eligible population comprises all individuals born between April 1st, 1979 and March 31st, 1980 who: (1) lived in a province included in the study continuously from their 12^th to their 22^nd birthday; (2) lived in an included province on their 12^th birthday but moved out of it before their 22^nd birthday, and (3) moved into an included province after their 12^th birthday and before their 22^nd birthday, or (4) made multiple moves between provinces after their 12^th birthday and before their 22^nd birthday.

As a result of net migration, the total population of the birth cohort in the part of Canada included in the study experienced a small but steady net growth between 1991 and 2000 — the period during which the individuals within the cohort moved from 12 years of age to 21 years of age. The cohort increased from 294,376 twelve year olds in 1991 to 323,694 twenty-one year olds in 2000. This represents an average annual increase of 1% or an overall increase of 10% in the size of the cohort.

Age-specific prevalence rates are calculated using yearly population data to determine the approximate population of males and females in each specific year for that corresponding age group. As such, changes in the population are not considered problematic because any gains or losses — through migration or death — are taken into account. However, when calculations of overall prevalence are concerned, the changing denominator (size of the total eligible population at risk) becomes problematic.

For purposes of estimating overall prevalence, the study utilizes the largest approximate population — the number of 21 year olds in 2000 — in its calculations. This approach accounts not only for the stable component of the original cohort size, but also the net growth experienced over time. Lee (1999) used a similar approach and rationale in determining the total eligible population in presenting overall prevalence estimates in a study of youth crime trends in British Columbia for four separate cohorts. An alternative, and less desirable, method uses the number of live births in the cohort birth year as an approximation of cohort size throughout the time period under study (see Prime et al., 2001 for an example of this use). This method would result in overall prevalence estimates approximately 3% higher than those produced by the method adopted in the present study.

Criminal charges heard in municipal courts in Quebec

During the period covered by this study, approximately 25% of Criminal Code charges involving adults in Quebec were heard in municipal courts, not provincial courts (Thomas, 2004: 11). Data for these charges were not captured by the Adult Criminal Court Survey, and consequently are not included in the present research. This section attempts to estimate how much the prevalence estimates for Quebec would change if charges heard in municipal courts were included.

The reported prevalence estimates are based on counts of the number of persons who had at least one charge referred to youth court or provincial criminal court, related to an incident occurring between the 12^th and 22^nd birthdays. Thus, persons who had a charge in youth court or provincial court would already be included in the prevalence estimates, regardless of whether they had any charges heard in municipal court. Therefore, addition of municipal court data would increase prevalence estimates only in relation to persons who had no youth court charges, and no provincial court charges, in their court careers.

Although the number of such persons is not known, it is possible to estimate the approximate maximum and minimum numbers, if certain assumptions are made. Assuming that 25% of adult offenders in Quebec had all of their charges heard in municipal court, then the inclusion of municipal court data would result in an increase of 33.3% (25/75) in the estimated number of adult offenders. However, it would not result in an increase of 33.3% in the prevalence estimate, because any adult offender who had charges heard in youth court would already be included in the prevalence counts. Table 6 shows that, among the cohort as a whole, 43% had referred incidents as adults but not as young persons. Assuming that this is also true in Quebec, the maximum increase in estimated prevalence in Quebec due to the addition of municipal court data would be (0.43)*(0.33) = 0.14, or 14% of the estimate based on youth court and provincial court data.

Table A2 gives prevalence estimates for referral to court and conviction in Quebec of 10.9 and 9.2 per 100 cohort members. Therefore, the maximum estimated prevalence of referral and conviction, adjusted to include municipal court data, would be 12.4 per 100 (10.9*1.14) and 10.5 per 100 (9.2*1.14) respectively. The minimum adjustment due to adding municipal court data would be zero, if all persons who had charges in municipal court were already included in the prevalence estimates because they also had charges in youth court or provincial court.

Therefore, the addition of data on criminal charges heard in municipal courts in Quebec would result in prevalence estimates between approximately 10.9 and 12.4 for referral to court, and between 9.2 and 10.5 for conviction.

Addition of municipal court data would also probably increase the estimated mean number of incidents in court careers in Quebec by a small amount, although we cannot be sure of this. The number of incidents in the court career would increase in the case of persons who had charges heard in municipal court, and who were already included in the study because they had charges heard in youth court or provincial court. However, an unknown number of additional persons would be added to the study population, who had all their charges heard in municipal court (thus increasing prevalence; see above), and who would probably have fewer incidents in their careers than the existing population (because they would necessarily not have any youth court charges), thus decreasing the mean number of incidents in the expanded population. Similarly, addition of data from the Quebec municipal courts would probably result in small changes in the estimates of other parameters of the court career, such as the annual rate of alleged offending, the duration of the career, etc.

Classification of offences and incidents

Classification of incidents

For most analyses involving breakdowns by the type of incident, incidents are classified into four groups — against the person, against property, against the administration of justice, and other — according to the nature of the most serious charge, or alleged offence, resulting from the incident, using the offence classification table below.

The most serious charge in the incident is determined by the seriousness scale, which was developed by Canadian Centre for Justice Statistics to rank the seriousness of criminal offences. It is based on the average length of prison sentence imposed on convicted charges between 1994/95 and 2000/01 in criminal court (Robinson, 2004: 10).

A “substantive” incident is defined as an incident which resulted in at least one substantive charge. A substantive charge is defined as a charge involving an alleged offence against the person, against property, or an “other” offence; i.e. not an offence against the administration of justice. Therefore, an alternative definition of a substantive incident is: an incident which resulted in at least one charge other than administrative charges. An “administrative” incident is defined as an incident which resulted solely in administrative charge(s).

Notes

1. This is fiscal 1979/80. The birth cohort was defined by a fiscal year rather than a calendar year in order to be consistent with the (fiscal) reporting period for the court data.

2. Approximately 25% of Criminal Code charges in Quebec are heard in municipal courts, not provincial courts (Thomas, 2004: 11). These charges are not captured by the ACCS, and consequently could not be included in the present study, which is therefore limited to incidents and persons processed in youth courts and provincial (criminal) courts. Although the impact on the results of this research of the omission of municipal court data for Quebec cannot be estimated precisely, it is not large; see the Methodology section, “Criminal charges heard in municipal courts in Quebec”.

3. For more information on record matching problems in the data for the Yukon, see footnote 7.

4. Data files were also constructed for the cohorts born in 1980/81 and 1981/82. These were used for validity checking of results obtained for the 1979/80 cohort. However, results from the two younger cohorts are not reported because their careers could be followed only until their 20th and 19th birthdays respectively.

5. The incident is certainly not the only possible intermediate level of aggregation of charge records. Others could be used, depending on the researcher’s interest. Research on sentencing would probably aggregate charges to the “case-at-disposition”; that is, the case consisting of all charges which were disposed of in one hearing. This is the aggregation of charges which is currently used in the reports by Canadian Centre for Justice Statistics on the court surveys (e.g. Ciccone McCutcheon, 2003; Robinson, 2004).

6. Data for the province of Quebec use the Henri code, which is more suited to coding French-Canadian names. See Armstrong (2000) for details of the Soundex and Henri codes, and a discussion of the issues surrounding their use in record matching.

7. Manual verification of name and date of birth information in the YCS and ACCS for a sample of records from the Yukon found an unacceptably high number of false negative matches, due in part to changes of name by residents of that territory. Therefore, the Yukon was omitted from the study. This problem was not apparent in the results of manual verification of data from the other jurisdictions.